{
 "meta": {
  "title": "Every AI that has run above the Kármán line, graded by evidence",
  "reviewed_by": "an AI session on behalf of the owner",
  "reviewed_on": "2026-10-04",
  "licence": "CC BY 4.0 for the register's own data. Quotations remain their sources'.",
  "boundary_km": 100
 },
 "records": [
  {
   "id": "remote-agent-ds1",
   "entry": 1,
   "name": "Remote Agent on Deep Space 1",
   "group": "ran",
   "class": "planning",
   "location": "deep space",
   "operator": "NASA Jet Propulsion Laboratory",
   "developer": "NASA",
   "host": "Deep Space 1",
   "catalogue": {
    "number": 25508,
    "name": "DEEP SPACE 1 (DS1)",
    "launched": "1998-10-24",
    "decayed": null,
    "match": "by name, the reviewer's judgement"
   },
   "dates": {
    "launched": "1998-10-24",
    "ran": "not dated on the page read",
    "reported": null,
    "ended": null
   },
   "hardware": "not stated in what was read",
   "software": "Remote Agent, an onboard planner and executive",
   "did": "Ran selected subsystems of the spacecraft from plans made on board, and dealt with four simulated faults; a bug interrupted the first experiment.",
   "marks": [
    {
     "stage": "L",
     "evidence": 3,
     "basis": [
      "c4"
     ]
    },
    {
     "stage": "R",
     "evidence": 1,
     "basis": [
      "c1",
      "c2",
      "c3"
     ]
    }
   ],
   "claims": [
    {
     "id": "c1",
     "kind": "quote",
     "text": "an artificial intelligence system was placed on board to plan and execute spacecraft activities",
     "speaker": "NASA JPL",
     "relation": "developer",
     "source": "jpl-ds1-remote-agent",
     "url": "https://www.jpl.nasa.gov/nmp/ds1/tech/autora.html",
     "source_title": "Deep Space 1: Advanced Technologies: Remote Agent",
     "publisher": "NASA Jet Propulsion Laboratory",
     "source_kind": "project-page",
     "published": "undated",
     "published_basis": "no date on the page",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in part: about the first 1,600 of 2,800 characters"
    },
    {
     "id": "c2",
     "kind": "quote",
     "text": "Injection of four (simulated) faults tested ... ability to resolve or work around different classes of problems, and in each case it devised the correct response.",
     "speaker": "NASA JPL",
     "relation": "developer",
     "source": "jpl-ds1-remote-agent",
     "url": "https://www.jpl.nasa.gov/nmp/ds1/tech/autora.html",
     "source_title": "Deep Space 1: Advanced Technologies: Remote Agent",
     "publisher": "NASA Jet Propulsion Laboratory",
     "source_kind": "project-page",
     "published": "undated",
     "published_basis": "no date on the page",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in part: about the first 1,600 of 2,800 characters"
    },
    {
     "id": "c3",
     "kind": "quote",
     "text": "A bug in the executive interrupted the first experiment",
     "speaker": "NASA JPL",
     "relation": "developer",
     "source": "jpl-ds1-remote-agent",
     "url": "https://www.jpl.nasa.gov/nmp/ds1/tech/autora.html",
     "source_title": "Deep Space 1: Advanced Technologies: Remote Agent",
     "publisher": "NASA Jet Propulsion Laboratory",
     "source_kind": "project-page",
     "published": "undated",
     "published_basis": "no date on the page",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in part: about the first 1,600 of 2,800 characters"
    },
    {
     "id": "c4",
     "kind": "data",
     "fields": {
      "OBJECT_NAME": "DEEP SPACE 1 (DS1)",
      "LAUNCH_DATE": "1998-10-24"
     },
     "speaker": "CelesTrak satellite catalogue",
     "relation": "catalogue",
     "source": "celestrak-25508",
     "url": "https://celestrak.org/satcat/records.php?CATNR=25508&FORMAT=json",
     "source_title": "Satellite catalogue record 25508",
     "publisher": "CelesTrak",
     "source_kind": "catalogue",
     "published": "undated",
     "published_basis": "a live catalogue record; the access date is the date that matters",
     "accessed": "2026-10-04",
     "copy_route": "json",
     "read": "in full"
    }
   ],
   "conflicts": [],
   "notes": "The page gives no date for the experiments. A survey paper dates them to 1999; that paper is secondary and is not used here.",
   "changed": "Launch confirmed from the catalogue.",
   "headline": "L3 R1"
  },
  {
   "id": "ase-eo1",
   "entry": 2,
   "name": "Autonomous Sciencecraft Experiment on Earth Observing-1",
   "group": "ran",
   "class": "planning",
   "location": "Earth orbit",
   "operator": "NASA",
   "developer": "NASA Jet Propulsion Laboratory",
   "host": "Earth Observing-1",
   "catalogue": {
    "number": 26619,
    "name": "EO-1",
    "launched": "2000-11-21",
    "decayed": null,
    "match": "by name, the reviewer's judgement"
   },
   "dates": {
    "launched": "2000-11-21",
    "ran": "2003 to 2017",
    "reported": null,
    "ended": "2017"
   },
   "hardware": "not stated in what was read",
   "software": "Onboard planning and execution with machine learning and pattern recognition",
   "did": "Chose what to send down and re-aimed the satellite at events it detected, over fourteen years of operation.",
   "marks": [
    {
     "stage": "L",
     "evidence": 3,
     "basis": [
      "c5"
     ]
    },
    {
     "stage": "U",
     "evidence": 1,
     "basis": [
      "c1",
      "c2",
      "c3"
     ]
    }
   ],
   "claims": [
    {
     "id": "c1",
     "kind": "quote",
     "text": "The Autonomous Sciencecraft Experiment (ASE) is operating onboard the Earth Observing-1 mission since 2003.",
     "speaker": "JPL Artificial Intelligence Group",
     "relation": "developer",
     "source": "jpl-ase",
     "url": "https://ai.jpl.nasa.gov/public/projects/ase/",
     "source_title": "ASE (Autonomous Sciencecraft Experiment)",
     "publisher": "JPL Artificial Intelligence Group",
     "source_kind": "project-page",
     "published": "undated",
     "published_basis": "no date on the page; its text mentions 2017",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in part: about the first 1,700 of 3,900 characters"
    },
    {
     "id": "c2",
     "kind": "quote",
     "text": "The ASE software uses onboard continuous planning, robust task and goal-based execution, and onboard machine learning and pattern recognition to radically increase science return by enabling intelligent downlink selection and autonomous retargeting.",
     "speaker": "JPL Artificial Intelligence Group",
     "relation": "developer",
     "source": "jpl-ase",
     "url": "https://ai.jpl.nasa.gov/public/projects/ase/",
     "source_title": "ASE (Autonomous Sciencecraft Experiment)",
     "publisher": "JPL Artificial Intelligence Group",
     "source_kind": "project-page",
     "published": "undated",
     "published_basis": "no date on the page; its text mentions 2017",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in part: about the first 1,700 of 3,900 characters"
    },
    {
     "id": "c3",
     "kind": "quote",
     "text": "Science Scenarios Executed: 1470 Positive Triggers: 257",
     "speaker": "JPL Artificial Intelligence Group",
     "relation": "developer",
     "source": "jpl-ase",
     "url": "https://ai.jpl.nasa.gov/public/projects/ase/",
     "source_title": "ASE (Autonomous Sciencecraft Experiment)",
     "publisher": "JPL Artificial Intelligence Group",
     "source_kind": "project-page",
     "published": "undated",
     "published_basis": "no date on the page; its text mentions 2017",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in part: about the first 1,700 of 3,900 characters"
    },
    {
     "id": "c4",
     "kind": "quote",
     "text": "EO-1 was decomissioned in Spring 2017.",
     "speaker": "JPL Artificial Intelligence Group",
     "relation": "developer",
     "source": "jpl-ase",
     "url": "https://ai.jpl.nasa.gov/public/projects/ase/",
     "source_title": "ASE (Autonomous Sciencecraft Experiment)",
     "publisher": "JPL Artificial Intelligence Group",
     "source_kind": "project-page",
     "published": "undated",
     "published_basis": "no date on the page; its text mentions 2017",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in part: about the first 1,700 of 3,900 characters"
    },
    {
     "id": "c5",
     "kind": "data",
     "fields": {
      "OBJECT_NAME": "EO-1",
      "LAUNCH_DATE": "2000-11-21"
     },
     "speaker": "CelesTrak satellite catalogue",
     "relation": "catalogue",
     "source": "celestrak-26619",
     "url": "https://celestrak.org/satcat/records.php?CATNR=26619&FORMAT=json",
     "source_title": "Satellite catalogue record 26619",
     "publisher": "CelesTrak",
     "source_kind": "catalogue",
     "published": "undated",
     "published_basis": "a live catalogue record; the access date is the date that matters",
     "accessed": "2026-10-04",
     "copy_route": "json",
     "read": "in full"
    }
   ],
   "conflicts": [],
   "notes": "Graded U, not R, because the page reports years of routine use with counts. It is still the developer's own page.",
   "changed": "Launch confirmed from the catalogue.",
   "headline": "L3 U1"
  },
  {
   "id": "aegis",
   "entry": 3,
   "name": "AEGIS on the Mars rovers Opportunity, Curiosity and Perseverance",
   "group": "ran",
   "class": "perception",
   "location": "surface of Mars",
   "operator": "NASA",
   "developer": "NASA Jet Propulsion Laboratory",
   "host": "Opportunity, Curiosity, Perseverance",
   "catalogue": null,
   "dates": {
    "launched": null,
    "ran": "2010 on Opportunity; routine on Curiosity since May 2016; on Perseverance from 2022",
    "reported": "2016-07-21",
    "ended": null
   },
   "hardware": "the rovers' own computers, not named",
   "software": "AEGIS, using the Rockster rock-detection algorithm",
   "did": "Finds rocks in the rover's camera images, ranks them against the scientists' priorities and points an instrument at the best, with nobody on Earth in the loop.",
   "marks": [
    {
     "stage": "U",
     "evidence": 1,
     "basis": [
      "c1",
      "c2",
      "c3",
      "c5"
     ]
    }
   ],
   "claims": [
    {
     "id": "c1",
     "kind": "quote",
     "text": "uses the Rockster algorithm to detect rock targets in NavCam imagery and rank them against science-team priorities",
     "speaker": "JPL Artificial Intelligence Group",
     "relation": "developer",
     "source": "jpl-aegis",
     "url": "https://ai.jpl.nasa.gov/public/projects/aegis/",
     "source_title": "AEGIS (Autonomous Exploration for Gathering Increased Science)",
     "publisher": "JPL Artificial Intelligence Group",
     "source_kind": "project-page",
     "published": "undated",
     "published_basis": "no date on the page; its text mentions 2022",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in part: about the first 1,700 of 8,900 characters"
    },
    {
     "id": "c2",
     "kind": "quote",
     "text": "Since May 2016 it has been in routine use on the MSL Curiosity rover, selecting targets for the ChemCam laser-induced breakdown spectrometer",
     "speaker": "JPL Artificial Intelligence Group",
     "relation": "developer",
     "source": "jpl-aegis",
     "url": "https://ai.jpl.nasa.gov/public/projects/aegis/",
     "source_title": "AEGIS (Autonomous Exploration for Gathering Increased Science)",
     "publisher": "JPL Artificial Intelligence Group",
     "source_kind": "project-page",
     "published": "undated",
     "published_basis": "no date on the page; its text mentions 2022",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in part: about the first 1,700 of 8,900 characters"
    },
    {
     "id": "c3",
     "kind": "quote",
     "text": "In 2022, AEGIS was deployed on the Mars 2020 Perseverance rover, where it autonomously targets the SuperCam",
     "speaker": "JPL Artificial Intelligence Group",
     "relation": "developer",
     "source": "jpl-aegis",
     "url": "https://ai.jpl.nasa.gov/public/projects/aegis/",
     "source_title": "AEGIS (Autonomous Exploration for Gathering Increased Science)",
     "publisher": "JPL Artificial Intelligence Group",
     "source_kind": "project-page",
     "published": "undated",
     "published_basis": "no date on the page; its text mentions 2022",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in part: about the first 1,700 of 8,900 characters"
    },
    {
     "id": "c4",
     "kind": "quote",
     "text": "the first time autonomous target selection is available for an instrument of this kind on any robotic planetary mission",
     "speaker": "NASA JPL",
     "relation": "developer",
     "source": "jpl-aegis-release-2016",
     "url": "https://www.jpl.nasa.gov/news/nasa-mars-rover-can-choose-laser-targets-on-its-own/",
     "source_title": "NASA Mars Rover Can Choose Laser Targets on Its Own",
     "publisher": "NASA Jet Propulsion Laboratory",
     "source_kind": "press-release",
     "published": "2016-07-21",
     "published_basis": "page metadata",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in part: about the first 1,500 of 5,100 characters"
    },
    {
     "id": "c5",
     "kind": "quote",
     "text": "AEGIS was first demonstrated on the MER rover Opportunity in 2010, where it selected targets for the Pancam imager",
     "speaker": "JPL Artificial Intelligence Group",
     "relation": "developer",
     "source": "jpl-aegis",
     "url": "https://ai.jpl.nasa.gov/public/projects/aegis/",
     "source_title": "AEGIS (Autonomous Exploration for Gathering Increased Science)",
     "publisher": "JPL Artificial Intelligence Group",
     "source_kind": "project-page",
     "published": "undated",
     "published_basis": "no date on the page; its text mentions 2022",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in part: about the first 1,700 of 8,900 characters"
    }
   ],
   "conflicts": [],
   "notes": "No launch mark is given: the claim is about software on rovers already on Mars, and no catalogue record applies.",
   "changed": null,
   "headline": "U1"
  },
  {
   "id": "ipex",
   "entry": null,
   "name": "IPEX, the Intelligent Payload Experiment",
   "group": "ran",
   "class": "perception",
   "location": "Earth orbit",
   "operator": "NASA Jet Propulsion Laboratory, with California Polytechnic State University",
   "developer": "NASA Jet Propulsion Laboratory",
   "host": "IPEX, a CubeSat",
   "catalogue": {
    "number": 39471,
    "name": "CP8 (IPEX)",
    "launched": "2013-12-06",
    "decayed": null,
    "match": "by name, the reviewer's judgement"
   },
   "dates": {
    "launched": "2013-12-06",
    "ran": "from December 2013; operations completed in 2015",
    "reported": null,
    "ended": "2015"
   },
   "hardware": "not stated in what was read",
   "software": "Random decision forests learned on the ground, and visual salience software",
   "did": "Classified its own images on board and picked out the regions worth sending down.",
   "marks": [
    {
     "stage": "L",
     "evidence": 3,
     "basis": [
      "c3"
     ]
    },
    {
     "stage": "R",
     "evidence": 1,
     "basis": [
      "c1",
      "c2"
     ]
    }
   ],
   "claims": [
    {
     "id": "c1",
     "kind": "quote",
     "text": "IPEX used machine-learned random decision forests to classify images onboard (to downlink classification maps) and computer vision visual salience software to extract interesting regions for downlink in acquired imagery.",
     "speaker": "JPL Artificial Intelligence Group",
     "relation": "developer",
     "source": "jpl-ipex",
     "url": "https://ai.jpl.nasa.gov/public/projects/ipex/",
     "source_title": "IPEX (Intelligent Payload Experiment)",
     "publisher": "JPL Artificial Intelligence Group",
     "source_kind": "project-page",
     "published": "undated",
     "published_basis": "no date on the page",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in part: four sentences of 5,700 characters"
    },
    {
     "id": "c2",
     "kind": "quote",
     "text": "IPEX launched in December 2013 for a six month mission, achieved its success criteria in three months, and flew for 14 months completing operations in 2015.",
     "speaker": "JPL Artificial Intelligence Group",
     "relation": "developer",
     "source": "jpl-ipex",
     "url": "https://ai.jpl.nasa.gov/public/projects/ipex/",
     "source_title": "IPEX (Intelligent Payload Experiment)",
     "publisher": "JPL Artificial Intelligence Group",
     "source_kind": "project-page",
     "published": "undated",
     "published_basis": "no date on the page",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in part: four sentences of 5,700 characters"
    },
    {
     "id": "c3",
     "kind": "data",
     "fields": {
      "OBJECT_NAME": "CP8 (IPEX)",
      "LAUNCH_DATE": "2013-12-06"
     },
     "speaker": "CelesTrak satellite catalogue",
     "relation": "catalogue",
     "source": "celestrak-39471",
     "url": "https://celestrak.org/satcat/records.php?CATNR=39471&FORMAT=json",
     "source_title": "Satellite catalogue record 39471 (CP8 (IPEX))",
     "publisher": "CelesTrak",
     "source_kind": "catalogue",
     "published": "undated",
     "published_basis": "a live catalogue record; dated by when it was read",
     "accessed": "2026-10-04",
     "copy_route": "json",
     "read": "in full"
    }
   ],
   "conflicts": [],
   "notes": "The classifier was learned on the ground and run in orbit. The mission team's 2015 paper was opened but not read far enough to quote, so the run rests on the project page.",
   "changed": "Added on 4 October 2026, in the second round of reading.",
   "headline": "L3 R1"
  },
  {
   "id": "phisat-1",
   "entry": 4,
   "name": "Φ-sat-1, flown on FSSCat",
   "group": "ran",
   "class": "perception",
   "location": "Earth orbit",
   "operator": "European Space Agency",
   "developer": "University of Pisa, Ubotica, cosine and Sinergise, with ESA",
   "host": "FSSCat",
   "catalogue": null,
   "dates": {
    "launched": "2020-09",
    "ran": "during the mission; no date given in what was read",
    "reported": "2021-11-04",
    "ended": null
   },
   "hardware": "Intel Movidius Myriad 2, on the Eyes of Things board, behind the HyperScout-2 imager",
   "software": "CloudScout, a convolutional neural network for cloud detection",
   "did": "Detected clouds in hyperspectral images on board, so that cloudy data need not be sent down.",
   "marks": [
    {
     "stage": "L",
     "evidence": 1,
     "basis": [
      "c9"
     ]
    },
    {
     "stage": "R",
     "evidence": 2,
     "basis": [
      "c4",
      "c5",
      "c6"
     ]
    }
   ],
   "claims": [
    {
     "id": "c1",
     "kind": "quote",
     "text": "is the first experiment to demonstrate on-board feature detection using hyperspectral Earth observation Convolutional Neural Networks",
     "speaker": "ESA Φ-lab",
     "relation": "operator",
     "source": "esa-philab-phisats",
     "url": "https://philab.esa.int/flagship-programmes/phi-sats-programme/",
     "source_title": "Φ-sats programme",
     "publisher": "ESA Φ-lab",
     "source_kind": "project-page",
     "published": "undated",
     "published_basis": "no date in the text; written before Φ-sat-2 was contracted",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in part: four excerpts of 3,100 characters"
    },
    {
     "id": "c2",
     "kind": "quote",
     "text": "has successfully delivered its first image and cloud mask",
     "speaker": "ESA Φ-lab",
     "relation": "operator",
     "source": "esa-philab-phisats",
     "url": "https://philab.esa.int/flagship-programmes/phi-sats-programme/",
     "source_title": "Φ-sats programme",
     "publisher": "ESA Φ-lab",
     "source_kind": "project-page",
     "published": "undated",
     "published_basis": "no date in the text; written before Φ-sat-2 was contracted",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in part: four excerpts of 3,100 characters"
    },
    {
     "id": "c3",
     "kind": "quote",
     "text": "represents the first ever on-board AI deep CNN inference on a dedicated chip",
     "speaker": "Giuffrida et al., the mission team",
     "relation": "mission-team",
     "source": "ieee-phisat1-paper",
     "url": "https://doi.org/10.1109/TGRS.2021.3125567",
     "source_title": "The Φ-Sat-1 Mission: The First On-Board Deep Neural Network Demonstrator for Satellite Earth Observation",
     "publisher": "IEEE Transactions on Geoscience and Remote Sensing (Giuffrida and twelve co-authors, the mission team)",
     "source_kind": "paper",
     "published": "2021-11-04",
     "published_basis": "date of publication printed in the paper",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in part: section V.D in full, the start of the conclusion, and about 20 excerpts of 70,500 characters"
    },
    {
     "id": "c4",
     "kind": "quote",
     "text": "NN inference with an exemplar TinyYOLO [36] model was continuously executed, with all inference outputs exactly matching the reference values",
     "speaker": "Giuffrida et al., the mission team",
     "relation": "mission-team",
     "source": "ieee-phisat1-paper",
     "url": "https://doi.org/10.1109/TGRS.2021.3125567",
     "source_title": "The Φ-Sat-1 Mission: The First On-Board Deep Neural Network Demonstrator for Satellite Earth Observation",
     "publisher": "IEEE Transactions on Geoscience and Remote Sensing (Giuffrida and twelve co-authors, the mission team)",
     "source_kind": "paper",
     "published": "2021-11-04",
     "published_basis": "date of publication printed in the paper",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in part: section V.D in full, the start of the conclusion, and about 20 excerpts of 70,500 characters"
    },
    {
     "id": "c5",
     "kind": "quote",
     "text": "for live images acquired from the HyperScout-2 sensor",
     "speaker": "Giuffrida et al., the mission team",
     "relation": "mission-team",
     "source": "ieee-phisat1-paper",
     "url": "https://doi.org/10.1109/TGRS.2021.3125567",
     "source_title": "The Φ-Sat-1 Mission: The First On-Board Deep Neural Network Demonstrator for Satellite Earth Observation",
     "publisher": "IEEE Transactions on Geoscience and Remote Sensing (Giuffrida and twelve co-authors, the mission team)",
     "source_kind": "paper",
     "published": "2021-11-04",
     "published_basis": "date of publication printed in the paper",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in part: section V.D in full, the start of the conclusion, and about 20 excerpts of 70,500 characters"
    },
    {
     "id": "c6",
     "kind": "quote",
     "text": "The performance of an on-board inference engine based on a machine learning (ML) algorithm for cloud detection was validated in flight. This is the first time such an experiment has been conducted in space.",
     "speaker": "Giuffrida et al., the mission team",
     "relation": "mission-team",
     "source": "ieee-phisat1-paper",
     "url": "https://doi.org/10.1109/TGRS.2021.3125567",
     "source_title": "The Φ-Sat-1 Mission: The First On-Board Deep Neural Network Demonstrator for Satellite Earth Observation",
     "publisher": "IEEE Transactions on Geoscience and Remote Sensing (Giuffrida and twelve co-authors, the mission team)",
     "source_kind": "paper",
     "published": "2021-11-04",
     "published_basis": "date of publication printed in the paper",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in part: section V.D in full, the start of the conclusion, and about 20 excerpts of 70,500 characters"
    },
    {
     "id": "c7",
     "kind": "quote",
     "text": "the first artificial intelligence (AI) technology carried onboard a European Earth observation mission",
     "speaker": "ESA, relayed by eoPortal",
     "relation": "operator-side-directory",
     "source": "eoportal-phisat",
     "url": "https://www.eoportal.org/satellite-missions/phisat-1",
     "source_title": "PhiSat-1 & -2 (Φ-Sat-1 & -2) Nanosatellite Mission",
     "publisher": "eoPortal (a mission directory run by ESA)",
     "source_kind": "directory",
     "published": "2026-08-09",
     "published_basis": "page metadata; the page is revised over time",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in part: about 30 excerpts of 65,900 characters"
    },
    {
     "id": "c8",
     "kind": "quote",
     "text": "The first satellite to run a deep neural network on board",
     "speaker": "The Consciousness AI, a third-party tracker",
     "relation": "tracker",
     "source": "consciousness-tracker",
     "url": "https://theconsciousness.ai/consciousness-to-orbit/tracker/",
     "source_title": "Orbital Compute Tracker Showing Every Satellite Running Computers in Space",
     "publisher": "The Consciousness AI",
     "source_kind": "tracker",
     "published": "undated",
     "published_basis": "no date on the page",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in full"
    },
    {
     "id": "c9",
     "kind": "quote",
     "text": "Launched in September 2020 and in August 2024 respectively",
     "speaker": "eoPortal, an ESA-run directory",
     "relation": "operator-side-directory",
     "source": "eoportal-phisat",
     "url": "https://www.eoportal.org/satellite-missions/phisat-1",
     "source_title": "PhiSat-1 & -2 (Φ-Sat-1 & -2) Nanosatellite Mission",
     "publisher": "eoPortal (a mission directory run by ESA)",
     "source_kind": "directory",
     "published": "2026-08-09",
     "published_basis": "page metadata; the page is revised over time",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in part: about 30 excerpts of 65,900 characters"
    },
    {
     "id": "c10",
     "kind": "quote",
     "text": "The EoT board features the Intel Movidius Myriad 2 vision processing unit (VPU)",
     "speaker": "Giuffrida et al., the mission team",
     "relation": "mission-team",
     "source": "ieee-phisat1-paper",
     "url": "https://doi.org/10.1109/TGRS.2021.3125567",
     "source_title": "The Φ-Sat-1 Mission: The First On-Board Deep Neural Network Demonstrator for Satellite Earth Observation",
     "publisher": "IEEE Transactions on Geoscience and Remote Sensing (Giuffrida and twelve co-authors, the mission team)",
     "source_kind": "paper",
     "published": "2021-11-04",
     "published_basis": "date of publication printed in the paper",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in part: section V.D in full, the start of the conclusion, and about 20 excerpts of 70,500 characters"
    }
   ],
   "conflicts": [
    {
     "summary": "The accuracy figure often quoted for this mission, 88.4%, is for a synthetic test set on the ground. No in-orbit accuracy figure was found in the parts of the paper read.",
     "status": "noted"
    }
   ],
   "notes": "The paper is peer reviewed, but its authors are the mission team, so it counts as their own technical account (evidence 2), not as independent confirmation. No catalogue object was found under the names tried.",
   "changed": "Run raised from R1 to R2 after the mission paper was read.",
   "headline": "L1 R2"
  },
  {
   "id": "ops-sat-orbitai",
   "entry": 5,
   "name": "OPS-SAT: models trained and run on board",
   "group": "ran",
   "class": "learning",
   "location": "Earth orbit",
   "operator": "European Space Agency",
   "developer": "The OPS-SAT flight control team (Labrèche and colleagues)",
   "host": "OPS-SAT",
   "catalogue": {
    "number": 44878,
    "name": "OPS-SAT",
    "launched": "2019-12-18",
    "decayed": "2024-05-22",
    "match": "by name, the reviewer's judgement"
   },
   "dates": {
    "launched": "2019-12-18",
    "ran": "by April 2021",
    "reported": "2021-04-19",
    "ended": "2024-05-22"
   },
   "hardware": "not stated in what was read",
   "software": "Online machine learning algorithms for training; TensorFlow Lite for inference; k-means for clustering",
   "did": "Trained a fault-detection model in flight from the satellite's own sensor stream, and classified and clustered its own images.",
   "marks": [
    {
     "stage": "L",
     "evidence": 3,
     "basis": [
      "c6"
     ]
    },
    {
     "stage": "R",
     "evidence": 2,
     "basis": [
      "c3",
      "c4",
      "c5"
     ]
    }
   ],
   "claims": [
    {
     "id": "c1",
     "kind": "quote",
     "text": "To our knowledge, this is the first time an artificial intelligence model is trained onboard a flying mission.",
     "speaker": "Georges Labrèche, experimenter",
     "relation": "experimenter",
     "source": "labreche-orbitai",
     "url": "https://georges.fyi/opssat/first-machine-learning-models-trained-in-space/",
     "source_title": "First Machine Learning Models Trained in Space",
     "publisher": "Georges Labrèche (personal site of the experiment's lead)",
     "source_kind": "blog",
     "published": "2021-04-19",
     "published_basis": "page metadata",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in full"
    },
    {
     "id": "c2",
     "kind": "quote",
     "text": "One of our models already gave us predictions with a whopping 89% balanced accuracy after only a few runs.",
     "speaker": "Georges Labrèche, experimenter",
     "relation": "experimenter",
     "source": "labreche-orbitai",
     "url": "https://georges.fyi/opssat/first-machine-learning-models-trained-in-space/",
     "source_title": "First Machine Learning Models Trained in Space",
     "publisher": "Georges Labrèche (personal site of the experiment's lead)",
     "source_kind": "blog",
     "published": "2021-04-19",
     "published_basis": "page metadata",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in full"
    },
    {
     "id": "c3",
     "kind": "quote",
     "text": "achieving balanced accuracies ranging from 85% to 99% from models trained with the Adagarad RDA, AROW, and NHERD online ML algorithms",
     "speaker": "Labrèche et al., the flight control team",
     "relation": "mission-team",
     "source": "openalex-opssat-abstract",
     "url": "https://doi.org/10.1109/AERO53065.2022.9843402",
     "source_title": "OPS-SAT Spacecraft Autonomy with TensorFlow Lite, Unsupervised Learning, and Online Machine Learning",
     "publisher": "2022 IEEE Aerospace Conference (Labrèche and six co-authors, the flight control team)",
     "source_kind": "abstract",
     "published": "2022-03-05",
     "published_basis": "publication date held by OpenAlex",
     "accessed": "2026-10-04",
     "copy_route": "openalex",
     "read": "abstract in full, as held by OpenAlex; the paper itself could not be reached"
    },
    {
     "id": "c4",
     "kind": "quote",
     "text": "The ability to train models in-flight with data generated on-board without human involvement is an exciting first",
     "speaker": "Labrèche et al., the flight control team",
     "relation": "mission-team",
     "source": "openalex-opssat-abstract",
     "url": "https://doi.org/10.1109/AERO53065.2022.9843402",
     "source_title": "OPS-SAT Spacecraft Autonomy with TensorFlow Lite, Unsupervised Learning, and Online Machine Learning",
     "publisher": "2022 IEEE Aerospace Conference (Labrèche and six co-authors, the flight control team)",
     "source_kind": "abstract",
     "published": "2022-03-05",
     "published_basis": "publication date held by OpenAlex",
     "accessed": "2026-10-04",
     "copy_route": "openalex",
     "read": "abstract in full, as held by OpenAlex; the paper itself could not be reached"
    },
    {
     "id": "c5",
     "kind": "quote",
     "text": "image classification with Convolutional Neural Network (CNN) model inferences using TensorFlow Lite",
     "speaker": "Labrèche et al., the flight control team",
     "relation": "mission-team",
     "source": "openalex-opssat-abstract",
     "url": "https://doi.org/10.1109/AERO53065.2022.9843402",
     "source_title": "OPS-SAT Spacecraft Autonomy with TensorFlow Lite, Unsupervised Learning, and Online Machine Learning",
     "publisher": "2022 IEEE Aerospace Conference (Labrèche and six co-authors, the flight control team)",
     "source_kind": "abstract",
     "published": "2022-03-05",
     "published_basis": "publication date held by OpenAlex",
     "accessed": "2026-10-04",
     "copy_route": "openalex",
     "read": "abstract in full, as held by OpenAlex; the paper itself could not be reached"
    },
    {
     "id": "c6",
     "kind": "data",
     "fields": {
      "OBJECT_NAME": "OPS-SAT",
      "LAUNCH_DATE": "2019-12-18",
      "DECAY_DATE": "2024-05-22"
     },
     "speaker": "CelesTrak satellite catalogue",
     "relation": "catalogue",
     "source": "celestrak-44878",
     "url": "https://celestrak.org/satcat/records.php?CATNR=44878&FORMAT=json",
     "source_title": "Satellite catalogue record 44878",
     "publisher": "CelesTrak",
     "source_kind": "catalogue",
     "published": "undated",
     "published_basis": "a live catalogue record; the access date is the date that matters",
     "accessed": "2026-10-04",
     "copy_route": "json",
     "read": "in full"
    }
   ],
   "conflicts": [],
   "notes": "Evidence 2 rests on the paper's abstract, read in full. The paper's body could not be reached.",
   "changed": "Run raised from R1 to R2 after the paper's abstract was read.",
   "headline": "L3 R2"
  },
  {
   "id": "worldfloods-wild-ride",
   "entry": null,
   "name": "WorldFloods on D-Orbit's Wild Ride mission",
   "group": "ran",
   "class": "perception",
   "location": "Earth orbit",
   "operator": "D-Orbit (the satellite)",
   "developer": "Trillium Technologies and the University of Oxford, with D-Orbit, Unibap and ESA Φ-lab",
   "host": "ION SCV Dauntless David",
   "catalogue": {
    "number": 48912,
    "name": "ION SCV-003",
    "launched": "2021-06-30",
    "decayed": "2024-12-08",
    "match": "by launch date: the only ION carrier in the catalogue launched on 30 June 2021, the date the paper gives; the reviewer's judgement"
   },
   "dates": {
    "launched": "2021-06-30",
    "ran": "after June 2021; not dated in the parts read",
    "reported": "2023-06-27",
    "ended": "2024-12-08"
   },
   "hardware": "A quad-core x86 processor with an Intel Movidius Myriad X",
   "software": "WorldFloods, a segmentation model that maps flood water",
   "did": "Produced flood maps on board from images, and was updated from the ground after launch to suit the satellite's own camera.",
   "marks": [
    {
     "stage": "L",
     "evidence": 3,
     "basis": [
      "c5"
     ]
    },
    {
     "stage": "R",
     "evidence": 2,
     "basis": [
      "c1",
      "c2"
     ]
    }
   ],
   "claims": [
    {
     "id": "c1",
     "kind": "quote",
     "text": "In this work we demonstrate, in a satellite deployed in orbit, a ML payload called 'WorldFloods' that is able to send compressed flood maps from sensed images.",
     "speaker": "Mateo-Garcia et al., the experiment's team",
     "relation": "experimenter",
     "source": "nature-worldfloods",
     "url": "https://www.nature.com/articles/s41598-023-34436-w",
     "source_title": "In-orbit demonstration of a re-trainable machine learning payload for processing optical imagery",
     "publisher": "Scientific Reports (Mateo-Garcia and co-authors, the experiment's team)",
     "source_kind": "paper",
     "published": "2023-06-27",
     "published_basis": "page metadata",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in part: about twenty sentences of 56,100 characters"
    },
    {
     "id": "c2",
     "kind": "quote",
     "text": "For the first time, we demonstrate a successful update of the ML model to adapt it to this camera after the satellite was launched.",
     "speaker": "Mateo-Garcia et al., the experiment's team",
     "relation": "experimenter",
     "source": "nature-worldfloods",
     "url": "https://www.nature.com/articles/s41598-023-34436-w",
     "source_title": "In-orbit demonstration of a re-trainable machine learning payload for processing optical imagery",
     "publisher": "Scientific Reports (Mateo-Garcia and co-authors, the experiment's team)",
     "source_kind": "paper",
     "published": "2023-06-27",
     "published_basis": "page metadata",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in part: about twenty sentences of 56,100 characters"
    },
    {
     "id": "c3",
     "kind": "quote",
     "text": "offering a quad-core x86 64-bit processor, a Microsemi SmartFusion2 FPGA and an Intel Movidius Myriad X Vision Processing Unit (VPU)",
     "speaker": "Mateo-Garcia et al., the experiment's team",
     "relation": "experimenter",
     "source": "nature-worldfloods",
     "url": "https://www.nature.com/articles/s41598-023-34436-w",
     "source_title": "In-orbit demonstration of a re-trainable machine learning payload for processing optical imagery",
     "publisher": "Scientific Reports (Mateo-Garcia and co-authors, the experiment's team)",
     "source_kind": "paper",
     "published": "2023-06-27",
     "published_basis": "page metadata",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in part: about twenty sentences of 56,100 characters"
    },
    {
     "id": "c4",
     "kind": "quote",
     "text": "The D-Orbit Wild Ride mission for the carrier ION SCV Dauntless David successfully launched into LEO on a SpaceX Falcon 9 rocket on June 30th 2021",
     "speaker": "Mateo-Garcia et al., the experiment's team",
     "relation": "experimenter",
     "source": "nature-worldfloods",
     "url": "https://www.nature.com/articles/s41598-023-34436-w",
     "source_title": "In-orbit demonstration of a re-trainable machine learning payload for processing optical imagery",
     "publisher": "Scientific Reports (Mateo-Garcia and co-authors, the experiment's team)",
     "source_kind": "paper",
     "published": "2023-06-27",
     "published_basis": "page metadata",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in part: about twenty sentences of 56,100 characters"
    },
    {
     "id": "c5",
     "kind": "data",
     "fields": {
      "OBJECT_NAME": "ION SCV-003",
      "LAUNCH_DATE": "2021-06-30",
      "DECAY_DATE": "2024-12-08"
     },
     "speaker": "CelesTrak satellite catalogue",
     "relation": "catalogue",
     "source": "celestrak-48912",
     "url": "https://celestrak.org/satcat/records.php?CATNR=48912&FORMAT=json",
     "source_title": "Satellite catalogue record 48912 (ION SCV-003)",
     "publisher": "CelesTrak",
     "source_kind": "catalogue",
     "published": "undated",
     "published_basis": "a live catalogue record; dated by when it was read",
     "accessed": "2026-10-04",
     "copy_route": "json",
     "read": "in full"
    }
   ],
   "conflicts": [],
   "notes": "Peer reviewed, and written by the people who ran the experiment, so it counts as their own technical account. The paper does not use the catalogue's name for the satellite; the match rests on the launch date.",
   "changed": "Added on 4 October 2026, in the second round of reading.",
   "headline": "L3 R2"
  },
  {
   "id": "aws-ion-scv004",
   "entry": null,
   "name": "AWS machine learning software on D-Orbit's ION SCV-004",
   "group": "ran",
   "class": "perception",
   "location": "Earth orbit",
   "operator": "D-Orbit (the satellite)",
   "developer": "Amazon Web Services, with D-Orbit and Unibap",
   "host": "ION SCV-004, Elysian Eleonora",
   "catalogue": {
    "number": 50993,
    "name": "ION SCV-004",
    "launched": "2022-01-13",
    "decayed": null,
    "match": "by name: the team's paper names the carrier SCV-004"
   },
   "dates": {
    "launched": "2022-01-13",
    "ran": "during 2022, after first contact on 21 January",
    "reported": "2022-11-29",
    "ended": null
   },
   "hardware": "Unibap SpaceCloud iX5",
   "software": "AWS edge computing and machine learning software; the models are not named in what was read",
   "did": "Ran inference on Earth observation data in orbit, and had its model settings changed in flight.",
   "marks": [
    {
     "stage": "L",
     "evidence": 3,
     "basis": [
      "c5"
     ]
    },
    {
     "stage": "R",
     "evidence": 2,
     "basis": [
      "c1",
      "c3"
     ]
    }
   ],
   "claims": [
    {
     "id": "c1",
     "kind": "quote",
     "text": "successfully ran a suite of AWS compute and machine learning (ML) software on an orbiting satellite, in a first-of-its-kind space experiment",
     "speaker": "Amazon Web Services",
     "relation": "developer",
     "source": "aws-orbit-ml",
     "url": "https://aws.amazon.com/blogs/publicsector/aws-successfully-runs-aws-compute-machine-learning-services-orbiting-satellite-first-space-experiment/",
     "source_title": "AWS successfully runs AWS compute and machine learning services on an orbiting satellite in a first-of-its kind space experiment",
     "publisher": "Amazon Web Services",
     "source_kind": "blog",
     "published": "2022-11-29",
     "published_basis": "date given in search results",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in part: about six sentences of 7,900 characters"
    },
    {
     "id": "c2",
     "kind": "quote",
     "text": "On January 21, 2022, the team made its first successful contact with the payload, and executed the first remote command from Earth to space.",
     "speaker": "Amazon Web Services",
     "relation": "developer",
     "source": "aws-orbit-ml",
     "url": "https://aws.amazon.com/blogs/publicsector/aws-successfully-runs-aws-compute-machine-learning-services-orbiting-satellite-first-space-experiment/",
     "source_title": "AWS successfully runs AWS compute and machine learning services on an orbiting satellite in a first-of-its kind space experiment",
     "publisher": "Amazon Web Services",
     "source_kind": "blog",
     "published": "2022-11-29",
     "published_basis": "date given in search results",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in part: about six sentences of 7,900 characters"
    },
    {
     "id": "c3",
     "kind": "quote",
     "text": "We successfully achieved the two primary goals of our experiment: perform inference and processing on data while in orbit using AWS edge capabilities, and change model configurations in flight using AWS edge service APIs.",
     "speaker": "O'Donnell et al., the experiment's team",
     "relation": "experimenter",
     "source": "smallsat-aws-cloud",
     "url": "https://digitalcommons.usu.edu/cgi/viewcontent.cgi?article=5659&context=smallsat",
     "source_title": "Extension of Cloud Computing to Small Satellites",
     "publisher": "Small Satellite Conference 2023 (O'Donnell and five co-authors at Amazon Web Services, the experiment's team)",
     "source_kind": "paper",
     "published": "2023",
     "published_basis": "the paper's number, SSC23-XIII-04",
     "accessed": "2026-10-04",
     "copy_route": "wayback",
     "read": "in part: about ten sentences of 22,800 characters, from an Internet Archive copy"
    },
    {
     "id": "c4",
     "kind": "quote",
     "text": "the D-Orbit ION Satellite Carrier SCV-004, Elysian Elenora",
     "speaker": "O'Donnell et al., the experiment's team",
     "relation": "experimenter",
     "source": "smallsat-aws-cloud",
     "url": "https://digitalcommons.usu.edu/cgi/viewcontent.cgi?article=5659&context=smallsat",
     "source_title": "Extension of Cloud Computing to Small Satellites",
     "publisher": "Small Satellite Conference 2023 (O'Donnell and five co-authors at Amazon Web Services, the experiment's team)",
     "source_kind": "paper",
     "published": "2023",
     "published_basis": "the paper's number, SSC23-XIII-04",
     "accessed": "2026-10-04",
     "copy_route": "wayback",
     "read": "in part: about ten sentences of 22,800 characters, from an Internet Archive copy"
    },
    {
     "id": "c5",
     "kind": "data",
     "fields": {
      "OBJECT_NAME": "ION SCV-004",
      "LAUNCH_DATE": "2022-01-13"
     },
     "speaker": "CelesTrak satellite catalogue",
     "relation": "catalogue",
     "source": "celestrak-50993",
     "url": "https://celestrak.org/satcat/records.php?CATNR=50993&FORMAT=json",
     "source_title": "Satellite catalogue record 50993 (ION SCV-004)",
     "publisher": "CelesTrak",
     "source_kind": "catalogue",
     "published": "undated",
     "published_basis": "a live catalogue record; dated by when it was read",
     "accessed": "2026-10-04",
     "copy_route": "json",
     "read": "in full"
    }
   ],
   "conflicts": [],
   "notes": "The claim to a first is scoped to the company's own software. What the models detected, and how well, is not in the parts read.",
   "changed": "Added on 4 October 2026, in the second round of reading.",
   "headline": "L3 R2"
  },
  {
   "id": "iss-ml-benchmark",
   "entry": null,
   "name": "Neural networks benchmarked on phone-class chips aboard the International Space Station",
   "group": "ran",
   "class": "perception",
   "location": "space station",
   "operator": "Hewlett Packard Enterprise (the host computer)",
   "developer": "NASA Jet Propulsion Laboratory, with Ubotica",
   "host": "HPE Spaceborne Computer-2, International Space Station",
   "catalogue": null,
   "dates": {
    "launched": null,
    "ran": "by 2022",
    "reported": "2022",
    "ended": null
   },
   "hardware": "Qualcomm Snapdragon 855 and Intel Movidius Myriad X, hosted by HPE Spaceborne Computer-2",
   "software": "Deep learning models for Mars rover imagery, spectral unmixing and ship detection, by the paper's title",
   "did": "Ran deep learning inference on two commercial processors aboard the station, to measure how they perform there.",
   "marks": [
    {
     "stage": "R",
     "evidence": 2,
     "basis": [
      "c1"
     ]
    }
   ],
   "claims": [
    {
     "id": "c1",
     "kind": "quote",
     "text": "We benchmark deep learning inference on Movidius Myriad X and Snapdragon processors onboard the ISS.",
     "speaker": "Dunkel et al., JPL and Ubotica, the experiment's team",
     "relation": "experimenter",
     "source": "jpl-iss-benchmark-paper",
     "url": "https://ai.jpl.nasa.gov/public/documents/papers/Dunkel-DL-ISS-ASTRA-2022.pdf",
     "source_title": "Testing Mars Rover, Spectral Unmixing, and Ship Detection Neural Networks, and Memory Checkers on Embedded Systems Onboard the ISS",
     "publisher": "Dunkel and co-authors, JPL and Ubotica (the experiment's team)",
     "source_kind": "paper",
     "published": "2022",
     "published_basis": "year in the file's address; no date read in the text",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in part: four sentences of 34,900 characters"
    },
    {
     "id": "c2",
     "kind": "quote",
     "text": "The Snapdragon 855 and Movidius are onboard the International Space Station supported by the Spaceborne Computing-2 by Hewlett Packard Enterprise.",
     "speaker": "JPL Artificial Intelligence Group",
     "relation": "developer",
     "source": "jpl-iss-benchmark",
     "url": "https://ai.jpl.nasa.gov/public/projects/iss/",
     "source_title": "Artificial Intelligence and Advanced Flight Computing on the International Space Station",
     "publisher": "JPL Artificial Intelligence Group",
     "source_kind": "project-page",
     "published": "undated",
     "published_basis": "no date on the page",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in part: four sentences of 4,100 characters"
    },
    {
     "id": "c3",
     "kind": "quote",
     "text": "We benchmark a suite of flight software (FSW) applications on the Qualcomm Snapdragon 855 handheld Development Kit (HDK), a high performance embedded processor used in mobile phones.",
     "speaker": "JPL Artificial Intelligence Group",
     "relation": "developer",
     "source": "jpl-iss-benchmark",
     "url": "https://ai.jpl.nasa.gov/public/projects/iss/",
     "source_title": "Artificial Intelligence and Advanced Flight Computing on the International Space Station",
     "publisher": "JPL Artificial Intelligence Group",
     "source_kind": "project-page",
     "published": "undated",
     "published_basis": "no date on the page",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in part: four sentences of 4,100 characters"
    }
   ],
   "conflicts": [],
   "notes": "The models were trained on the ground and their inputs were prepared in advance; this is a test of processors, not of a model at work. The release about Dynamic Targeting says the same team ran similar algorithms on the station in 2022, which this appears to be.",
   "changed": "Added on 4 October 2026, in the second round of reading.",
   "headline": "R2"
  },
  {
   "id": "dart-smart-nav",
   "entry": null,
   "name": "DART's autonomous targeting",
   "group": "ran",
   "class": "classical",
   "location": "deep space",
   "operator": "Johns Hopkins Applied Physics Laboratory, for NASA",
   "developer": "Johns Hopkins Applied Physics Laboratory",
   "host": "DART",
   "catalogue": {
    "number": 49497,
    "name": "NASA DART",
    "launched": "2021-11-24",
    "decayed": "2022-09-26",
    "match": "by name, the reviewer's judgement"
   },
   "dates": {
    "launched": "2021-11-24",
    "ran": "26 September 2022",
    "reported": "2023-09-26",
    "ended": "2022-09-26"
   },
   "hardware": "not stated in what was read",
   "software": "An autonomous targeting system, known as SMART Nav",
   "did": "Aimed the spacecraft at a small asteroid by itself, and hit it.",
   "marks": [
    {
     "stage": "L",
     "evidence": 3,
     "basis": [
      "c3"
     ]
    },
    {
     "stage": "R",
     "evidence": 1,
     "basis": [
      "c1",
      "c2"
     ]
    }
   ],
   "claims": [
    {
     "id": "c1",
     "kind": "quote",
     "text": "DART (Double Asteroid Redirection Test) used an autonomous targeting system to aim itself at Dimorphos.",
     "speaker": "Johns Hopkins Applied Physics Laboratory",
     "relation": "developer",
     "source": "apl-dart",
     "url": "https://www.jhuapl.edu/destinations/missions/dart",
     "source_title": "DART",
     "publisher": "Johns Hopkins Applied Physics Laboratory",
     "source_kind": "project-page",
     "published": "2023-09-26",
     "published_basis": "page metadata",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in part: three sentences of 5,300 characters"
    },
    {
     "id": "c2",
     "kind": "quote",
     "text": "DART impacted Dimorphos on Sept. 26, 2022, striking the smaller body at just over four miles per second",
     "speaker": "Johns Hopkins Applied Physics Laboratory",
     "relation": "developer",
     "source": "apl-dart",
     "url": "https://www.jhuapl.edu/destinations/missions/dart",
     "source_title": "DART",
     "publisher": "Johns Hopkins Applied Physics Laboratory",
     "source_kind": "project-page",
     "published": "2023-09-26",
     "published_basis": "page metadata",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in part: three sentences of 5,300 characters"
    },
    {
     "id": "c3",
     "kind": "data",
     "fields": {
      "OBJECT_NAME": "NASA DART",
      "LAUNCH_DATE": "2021-11-24",
      "DECAY_DATE": "2022-09-26"
     },
     "speaker": "CelesTrak satellite catalogue",
     "relation": "catalogue",
     "source": "celestrak-49497",
     "url": "https://celestrak.org/satcat/records.php?CATNR=49497&FORMAT=json",
     "source_title": "Satellite catalogue record 49497 (NASA DART)",
     "publisher": "CelesTrak",
     "source_kind": "catalogue",
     "published": "undated",
     "published_basis": "a live catalogue record; dated by when it was read",
     "accessed": "2026-10-04",
     "copy_route": "json",
     "read": "in full"
    }
   ],
   "conflicts": [],
   "notes": "Classed as classical autonomy: navigation from images by fixed algorithms, with no learned model named in what was read. The catalogue's decay date is the day of the impact.",
   "changed": "Added on 4 October 2026, in the second round of reading.",
   "headline": "L3 R1"
  },
  {
   "id": "starling-dsa",
   "entry": null,
   "name": "Distributed Spacecraft Autonomy on the Starling swarm",
   "group": "ran",
   "class": "classical",
   "location": "Earth orbit",
   "operator": "NASA Ames Research Center",
   "developer": "NASA Ames Research Center",
   "host": "Four Starling CubeSats",
   "catalogue": {
    "number": 57388,
    "name": "STARLING 1",
    "launched": "2023-07-18",
    "decayed": null,
    "match": "by name; one of the swarm's four satellites, the reviewer's judgement"
   },
   "dates": {
    "launched": "2023-07-18",
    "ran": "after July 2023; not dated in the parts read",
    "reported": "2024",
    "ended": null
   },
   "hardware": "not stated in what was read",
   "software": "Distributed planning software that shares status between spacecraft and plans observations by optimisation",
   "did": "Let four satellites decide between themselves, with no instruction from the ground, which signals each should observe.",
   "marks": [
    {
     "stage": "L",
     "evidence": 3,
     "basis": [
      "c4"
     ]
    },
    {
     "stage": "R",
     "evidence": 2,
     "basis": [
      "c1",
      "c2",
      "c3"
     ]
    }
   ],
   "claims": [
    {
     "id": "c1",
     "kind": "quote",
     "text": "The accomplishments of DSA onboard Starling include the first fully distributed autonomous operation of multiple spacecraft, the first use of space-to-space communications to autonomously share status information between multiple spacecraft",
     "speaker": "NASA",
     "relation": "developer",
     "source": "nasa-starling",
     "url": "https://www.nasa.gov/smallspacecraft/what-is-starling/",
     "source_title": "What is Starling?",
     "publisher": "NASA",
     "source_kind": "project-page",
     "published": "undated",
     "published_basis": "page metadata gives 19 July 2023, the week of launch, but the text reports later results, so the page has been revised since",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in part: five sentences of 10,700 characters"
    },
    {
     "id": "c2",
     "kind": "quote",
     "text": "DSA onboard software autonomously coordinated the selection of the best GPS signals across all Starling spacecraft to accurately capture regions of higher or lower electron density.",
     "speaker": "NASA",
     "relation": "developer",
     "source": "nasa-starling",
     "url": "https://www.nasa.gov/smallspacecraft/what-is-starling/",
     "source_title": "What is Starling?",
     "publisher": "NASA",
     "source_kind": "project-page",
     "published": "undated",
     "published_basis": "page metadata gives 19 July 2023, the week of launch, but the text reports later results, so the page has been revised since",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in part: five sentences of 10,700 characters"
    },
    {
     "id": "c3",
     "kind": "quote",
     "text": "The DSA mission successfully ran on orbit and results were delivered.",
     "speaker": "Adams et al., NASA Ames, the experiment's team",
     "relation": "experimenter",
     "source": "smallsat-dsa-results",
     "url": "https://digitalcommons.usu.edu/cgi/viewcontent.cgi?article=5837&context=smallsat",
     "source_title": "Results From the Distributed Spacecraft Autonomy Experiment on Starling",
     "publisher": "Small Satellite Conference 2024 (Adams and co-authors, NASA Ames, the experiment's team)",
     "source_kind": "paper",
     "published": "2024",
     "published_basis": "year given in search results",
     "accessed": "2026-10-04",
     "copy_route": "wayback",
     "read": "in part: four sentences of 40,100 characters"
    },
    {
     "id": "c4",
     "kind": "data",
     "fields": {
      "OBJECT_NAME": "STARLING 1",
      "LAUNCH_DATE": "2023-07-18"
     },
     "speaker": "CelesTrak satellite catalogue",
     "relation": "catalogue",
     "source": "celestrak-57388",
     "url": "https://celestrak.org/satcat/records.php?CATNR=57388&FORMAT=json",
     "source_title": "Satellite catalogue record 57388 (STARLING 1)",
     "publisher": "CelesTrak",
     "source_kind": "catalogue",
     "published": "undated",
     "published_basis": "a live catalogue record; dated by when it was read",
     "accessed": "2026-10-04",
     "copy_route": "json",
     "read": "in full"
    }
   ],
   "conflicts": [
    {
     "summary": "The launch date. NASA's page gives 17 July 2023; the catalogue gives 18 July 2023. The catalogue keeps universal time, which would account for a day.",
     "status": "noted; the catalogue's date is used"
    }
   ],
   "notes": "Classed as classical autonomy: the planning is done by optimisation, with no learned model named in what was read. It is in the register and out of the count of models run.",
   "changed": "Added on 4 October 2026, in the second round of reading.",
   "headline": "L3 R2"
  },
  {
   "id": "fluid-iss",
   "entry": 6,
   "name": "Project FLUID: federated learning on the International Space Station",
   "group": "ran",
   "class": "learning",
   "location": "space station",
   "operator": "Hewlett Packard Enterprise, with NASA",
   "developer": "Hewlett Packard Enterprise",
   "host": "HPE Spaceborne Computer-2, International Space Station",
   "catalogue": null,
   "dates": {
    "launched": null,
    "ran": "by February 2024",
    "reported": "2024-02-29",
    "ended": null
   },
   "hardware": "HPE Spaceborne Computer-2",
   "software": "Federated learning; no model named",
   "did": "Trained machine learning models jointly from data held on Earth and data held on the station, without pooling the data.",
   "marks": [
    {
     "stage": "R",
     "evidence": 1,
     "basis": [
      "c1",
      "c2"
     ]
    }
   ],
   "claims": [
    {
     "id": "c1",
     "kind": "quote",
     "text": "HPE completes first Project FLUID experiment in collaboration with NASA",
     "speaker": "Hewlett Packard Enterprise",
     "relation": "host-operator",
     "source": "hpe-spaceborne-timeline",
     "published": "2024-02-29",
     "published_basis": "date of the entry on the page",
     "url": "https://www.hpe.com/us/en/newsroom/accelerating-space-exploration-with-the-spaceborne-computer.html",
     "source_title": "Accelerating space exploration with the Spaceborne Computer",
     "publisher": "Hewlett Packard Enterprise",
     "source_kind": "timeline",
     "accessed": "2026-10-04",
     "copy_route": "manual",
     "read": "in part: the first 16 of 64 timeline entries, and four entries in full"
    },
    {
     "id": "c2",
     "kind": "quote",
     "text": "This experiment resulted in the first successful implementation of machine learning models trained in a federated manner using data held separately on both Earth and aboard the International Space Station.",
     "speaker": "Hewlett Packard Enterprise",
     "relation": "host-operator",
     "source": "hpe-spaceborne-timeline",
     "published": "2024-02-29",
     "published_basis": "date of the entry on the page",
     "url": "https://www.hpe.com/us/en/newsroom/accelerating-space-exploration-with-the-spaceborne-computer.html",
     "source_title": "Accelerating space exploration with the Spaceborne Computer",
     "publisher": "Hewlett Packard Enterprise",
     "source_kind": "timeline",
     "accessed": "2026-10-04",
     "copy_route": "manual",
     "read": "in part: the first 16 of 64 timeline entries, and four entries in full"
    }
   ],
   "conflicts": [],
   "notes": "The source is a timeline entry of two sentences. No paper or data was found.",
   "changed": null,
   "headline": "R1"
  },
  {
   "id": "pixl-adaptive-sampling",
   "entry": null,
   "name": "Adaptive sampling on Perseverance's PIXL instrument",
   "group": "ran",
   "class": "perception",
   "location": "surface of Mars",
   "operator": "NASA",
   "developer": "NASA Jet Propulsion Laboratory",
   "host": "Perseverance",
   "catalogue": null,
   "dates": {
    "launched": null,
    "ran": "for almost three years to July 2024",
    "reported": "2024-07-16",
    "ended": null
   },
   "hardware": "the rover's own computer, not named",
   "software": "Adaptive sampling software for the PIXL X-ray instrument",
   "did": "Read the instrument's scans of a rock as they came in and decided, on Mars, where to look longer.",
   "marks": [
    {
     "stage": "U",
     "evidence": 1,
     "basis": [
      "c1",
      "c2",
      "c3"
     ]
    }
   ],
   "claims": [
    {
     "id": "c1",
     "kind": "quote",
     "text": "For almost three years, the rover mission has been testing a form of artificial intelligence that seeks out minerals in the Red Planet's rocks.",
     "speaker": "NASA JPL",
     "relation": "operator",
     "source": "nasa-pixl-adaptive-sampling",
     "url": "https://www.nasa.gov/missions/mars-2020-perseverance/perseverance-rover/heres-how-ai-is-changing-nasas-mars-rover-science/",
     "source_title": "Here's How AI Is Changing NASA's Mars Rover Science",
     "publisher": "NASA Jet Propulsion Laboratory",
     "source_kind": "press-release",
     "published": "2024-07-16",
     "published_basis": "page metadata",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in part: about seven sentences of 7,600 characters"
    },
    {
     "id": "c2",
     "kind": "quote",
     "text": "This marks the first time AI has been used on Mars to make autonomous decisions based on real-time analysis of rock composition.",
     "speaker": "NASA JPL",
     "relation": "operator",
     "source": "nasa-pixl-adaptive-sampling",
     "url": "https://www.nasa.gov/missions/mars-2020-perseverance/perseverance-rover/heres-how-ai-is-changing-nasas-mars-rover-science/",
     "source_title": "Here's How AI Is Changing NASA's Mars Rover Science",
     "publisher": "NASA Jet Propulsion Laboratory",
     "source_kind": "press-release",
     "published": "2024-07-16",
     "published_basis": "page metadata",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in part: about seven sentences of 7,600 characters"
    },
    {
     "id": "c3",
     "kind": "quote",
     "text": "the software autonomously positions the instrument close to a rock target, then looks at PIXL's scans of the target to find minerals worth examining more deeply",
     "speaker": "NASA JPL",
     "relation": "operator",
     "source": "nasa-pixl-adaptive-sampling",
     "url": "https://www.nasa.gov/missions/mars-2020-perseverance/perseverance-rover/heres-how-ai-is-changing-nasas-mars-rover-science/",
     "source_title": "Here's How AI Is Changing NASA's Mars Rover Science",
     "publisher": "NASA Jet Propulsion Laboratory",
     "source_kind": "press-release",
     "published": "2024-07-16",
     "published_basis": "page metadata",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in part: about seven sentences of 7,600 characters"
    }
   ],
   "conflicts": [],
   "notes": "Graded U, not R, because the release reports use over years. No launch mark is given: the claim is about software on a rover already on Mars.",
   "changed": "Added on 4 October 2026, in the second round of reading.",
   "headline": "U1"
  },
  {
   "id": "llm-iss",
   "entry": 7,
   "name": "A large language model on the International Space Station",
   "group": "ran",
   "class": "language",
   "location": "space station",
   "operator": "Hewlett Packard Enterprise",
   "developer": "Booz Allen Hamilton",
   "host": "HPE Spaceborne Computer-2, International Space Station",
   "catalogue": null,
   "dates": {
    "launched": null,
    "ran": "from mid-July 2024, by a news report of the developer's announcement",
    "reported": "2024-08-01",
    "ended": null
   },
   "hardware": "HPE Spaceborne Computer-2",
   "software": "A generative large language model with retrieval; not named",
   "did": "Answered queries from documents held on board, without a link to Earth.",
   "marks": [
    {
     "stage": "R",
     "evidence": 1,
     "basis": [
      "c1",
      "c3"
     ]
    }
   ],
   "claims": [
    {
     "id": "c1",
     "kind": "quote",
     "text": "the successful deployment and operation of a generative AI large language model (LLM) in space using Hewlett Packard Enterprise's (HPE) Spaceborne Computer-2 onboard the International Space Station (ISS) National Lab",
     "speaker": "Booz Allen Hamilton",
     "relation": "developer",
     "source": "boozallen-release",
     "url": "https://investors.boozallen.com/news-releases/news-release-details/booz-allen-deploys-power-generative-ai-space",
     "source_title": "Booz Allen Deploys the Power of Generative AI in Space",
     "publisher": "Booz Allen Hamilton",
     "source_kind": "press-release",
     "published": "2024-08-01",
     "published_basis": "dateline of the release",
     "accessed": "2026-10-04",
     "copy_route": "wayback",
     "read": "in part: about ten excerpts of 7,900 characters, from an Internet Archive capture of 10 Jul 2025"
    },
    {
     "id": "c2",
     "kind": "quote",
     "text": "This LLM is believed to be the first one deployed in space",
     "speaker": "Booz Allen Hamilton",
     "relation": "developer",
     "source": "boozallen-release",
     "url": "https://investors.boozallen.com/news-releases/news-release-details/booz-allen-deploys-power-generative-ai-space",
     "source_title": "Booz Allen Deploys the Power of Generative AI in Space",
     "publisher": "Booz Allen Hamilton",
     "source_kind": "press-release",
     "published": "2024-08-01",
     "published_basis": "dateline of the release",
     "accessed": "2026-10-04",
     "copy_route": "wayback",
     "read": "in part: about ten excerpts of 7,900 characters, from an Internet Archive capture of 10 Jul 2025"
    },
    {
     "id": "c3",
     "kind": "quote",
     "text": "HPE successfully deploys and operates genAI LLM on ISS",
     "speaker": "Hewlett Packard Enterprise",
     "relation": "host-operator",
     "source": "hpe-spaceborne-timeline",
     "published": "2024-08-01",
     "published_basis": "date of the entry on the page",
     "url": "https://www.hpe.com/us/en/newsroom/accelerating-space-exploration-with-the-spaceborne-computer.html",
     "source_title": "Accelerating space exploration with the Spaceborne Computer",
     "publisher": "Hewlett Packard Enterprise",
     "source_kind": "timeline",
     "accessed": "2026-10-04",
     "copy_route": "manual",
     "read": "in part: the first 16 of 64 timeline entries, and four entries in full"
    },
    {
     "id": "c4",
     "kind": "quote",
     "text": "This generative AI large language model (LLM) is believed to be the first LLM deployed in space",
     "speaker": "Hewlett Packard Enterprise",
     "relation": "host-operator",
     "source": "hpe-spaceborne-timeline",
     "published": "2024-08-01",
     "published_basis": "date of the entry on the page",
     "url": "https://www.hpe.com/us/en/newsroom/accelerating-space-exploration-with-the-spaceborne-computer.html",
     "source_title": "Accelerating space exploration with the Spaceborne Computer",
     "publisher": "Hewlett Packard Enterprise",
     "source_kind": "timeline",
     "accessed": "2026-10-04",
     "copy_route": "manual",
     "read": "in part: the first 16 of 64 timeline entries, and four entries in full"
    },
    {
     "id": "c5",
     "kind": "quote",
     "text": "retrieval-augmented generation",
     "speaker": "Booz Allen Hamilton",
     "relation": "developer",
     "source": "boozallen-release",
     "url": "https://investors.boozallen.com/news-releases/news-release-details/booz-allen-deploys-power-generative-ai-space",
     "source_title": "Booz Allen Deploys the Power of Generative AI in Space",
     "publisher": "Booz Allen Hamilton",
     "source_kind": "press-release",
     "published": "2024-08-01",
     "published_basis": "dateline of the release",
     "accessed": "2026-10-04",
     "copy_route": "wayback",
     "read": "in part: about ten excerpts of 7,900 characters, from an Internet Archive capture of 10 Jul 2025"
    },
    {
     "id": "c6",
     "kind": "quote",
     "text": "The generative AI large language model (LLM) has been in operation since mid-July as part of an experiment, Booz Allen announced Aug. 1.",
     "speaker": "SpaceNews",
     "relation": "news",
     "source": "spacenews-boozallen",
     "url": "https://spacenews.com/booz-allen-deploys-advanced-language-model-in-space/",
     "source_title": "Booz Allen deploys advanced language model in space",
     "publisher": "SpaceNews",
     "source_kind": "news",
     "published": "2024-08-01",
     "published_basis": "page metadata",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in part: one sentence of 2,100 characters; the rest is behind a registration wall"
    }
   ],
   "conflicts": [],
   "notes": "Two parties state this, the developer and the operator of the host computer. It is not counted as confirmed, because the host operator presents the deployment as its own and sells the computer.",
   "changed": "Booz Allen's release, which could not be opened before, was read from an archive capture. The mark is unchanged. A news report, which relays the developer, was added later the same day.",
   "headline": "R1"
  },
  {
   "id": "phisat-2",
   "entry": 8,
   "name": "Φ-sat-2",
   "group": "ran",
   "class": "perception",
   "location": "Earth orbit",
   "operator": "European Space Agency",
   "developer": "The Φ-sat-2 mission team",
   "host": "Φ-sat-2",
   "catalogue": {
    "number": 60470,
    "name": "PHISAT 2",
    "launched": "2024-08-16",
    "decayed": null,
    "match": "by name, the reviewer's judgement"
   },
   "dates": {
    "launched": "2024-08-16",
    "ran": "during commissioning, after August 2024",
    "reported": "2025-08-11",
    "ended": null
   },
   "hardware": "an \"AI accelerator\", not named in the abstract; ESA's earlier page planned an Intel Movidius Myriad 2",
   "software": "Six AI applications: deep compression, and detection of clouds, streets, vessels, fires and marine anomalies",
   "did": "Produced the outputs of its AI applications on board during commissioning, for comparison with results obtained on the ground.",
   "marks": [
    {
     "stage": "L",
     "evidence": 3,
     "basis": [
      "c5"
     ]
    },
    {
     "stage": "R",
     "evidence": 2,
     "basis": [
      "c2",
      "c3"
     ]
    }
   ],
   "claims": [
    {
     "id": "c1",
     "kind": "quote",
     "text": "Launched on August 16, 2024, aboard the SpaceX Transporter 11 rideshare flight",
     "speaker": "Melega et al., the mission team",
     "relation": "mission-team",
     "source": "smallsat-phisat2",
     "url": "https://digitalcommons.usu.edu/smallsat/2025/E-S1-2025/3/",
     "source_title": "In-Orbit Commissioning Results of the ESA Φsat-2 Mission",
     "publisher": "Small Satellite Conference 2025 (Melega and seven co-authors, the mission team)",
     "source_kind": "abstract",
     "published": "2025-08-11",
     "published_basis": "page metadata",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in part: about 2,600 of the abstract's 7,100 characters"
    },
    {
     "id": "c2",
     "kind": "quote",
     "text": "fine-tune six onboard AI applications, which include deep compression, cloud detection, street detection, vessel detection, fire detection, and marine anomalies detection",
     "speaker": "Melega et al., the mission team",
     "relation": "mission-team",
     "source": "smallsat-phisat2",
     "url": "https://digitalcommons.usu.edu/smallsat/2025/E-S1-2025/3/",
     "source_title": "In-Orbit Commissioning Results of the ESA Φsat-2 Mission",
     "publisher": "Small Satellite Conference 2025 (Melega and seven co-authors, the mission team)",
     "source_kind": "abstract",
     "published": "2025-08-11",
     "published_basis": "page metadata",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in part: about 2,600 of the abstract's 7,100 characters"
    },
    {
     "id": "c3",
     "kind": "quote",
     "text": "compare the abilities of the AI applications to match the results obtained with synthetic data",
     "speaker": "Melega et al., the mission team",
     "relation": "mission-team",
     "source": "smallsat-phisat2",
     "url": "https://digitalcommons.usu.edu/smallsat/2025/E-S1-2025/3/",
     "source_title": "In-Orbit Commissioning Results of the ESA Φsat-2 Mission",
     "publisher": "Small Satellite Conference 2025 (Melega and seven co-authors, the mission team)",
     "source_kind": "abstract",
     "published": "2025-08-11",
     "published_basis": "page metadata",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in part: about 2,600 of the abstract's 7,100 characters"
    },
    {
     "id": "c4",
     "kind": "quote",
     "text": "among the first to implement a complete data processing chain onboard",
     "speaker": "Melega et al., the mission team",
     "relation": "mission-team",
     "source": "smallsat-phisat2",
     "url": "https://digitalcommons.usu.edu/smallsat/2025/E-S1-2025/3/",
     "source_title": "In-Orbit Commissioning Results of the ESA Φsat-2 Mission",
     "publisher": "Small Satellite Conference 2025 (Melega and seven co-authors, the mission team)",
     "source_kind": "abstract",
     "published": "2025-08-11",
     "published_basis": "page metadata",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in part: about 2,600 of the abstract's 7,100 characters"
    },
    {
     "id": "c5",
     "kind": "data",
     "fields": {
      "OBJECT_NAME": "PHISAT 2",
      "LAUNCH_DATE": "2024-08-16"
     },
     "speaker": "CelesTrak satellite catalogue",
     "relation": "catalogue",
     "source": "celestrak-60470",
     "url": "https://celestrak.org/satcat/records.php?CATNR=60470&FORMAT=json",
     "source_title": "Satellite catalogue record 60470",
     "publisher": "CelesTrak",
     "source_kind": "catalogue",
     "published": "undated",
     "published_basis": "a live catalogue record; the access date is the date that matters",
     "accessed": "2026-10-04",
     "copy_route": "json",
     "read": "in full"
    },
    {
     "id": "c6",
     "kind": "quote",
     "text": "The payload will include the AI processor Intel Movidius Myriad 2",
     "speaker": "ESA Φ-lab",
     "relation": "operator",
     "source": "esa-philab-phisats",
     "url": "https://philab.esa.int/flagship-programmes/phi-sats-programme/",
     "source_title": "Φ-sats programme",
     "publisher": "ESA Φ-lab",
     "source_kind": "project-page",
     "published": "undated",
     "published_basis": "no date in the text; written before Φ-sat-2 was contracted",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in part: four excerpts of 3,100 characters"
    }
   ],
   "conflicts": [
    {
     "summary": "The launch date. An ESA-run directory, eoPortal, gives 16 June 2024 in two places and August 2024 in a third. The mission team's paper and the catalogue both give 16 August 2024.",
     "status": "resolved: 16 August 2024"
    }
   ],
   "notes": "The abstract reports that outputs were acquired and compared. It gives no accuracy for the applications in orbit.",
   "changed": "Launch confirmed from the catalogue.",
   "headline": "L3 R2"
  },
  {
   "id": "intuition-1",
   "entry": 9,
   "name": "Intuition-1",
   "group": "ran",
   "class": "perception",
   "location": "Earth orbit",
   "operator": "KP Labs",
   "developer": "KP Labs",
   "host": "Intuition-1",
   "catalogue": {
    "number": 58565,
    "name": "INTUITION-1",
    "launched": "2023-11-11",
    "decayed": "2026-03-23",
    "match": "by name, the reviewer's judgement"
   },
   "dates": {
    "launched": "2023-11-11",
    "ran": "not dated on the page read",
    "reported": null,
    "ended": "2026-03-23"
   },
   "hardware": "Leopard Data Processing Unit, with a 192-band hyperspectral sensor",
   "software": "not named",
   "did": "Processed its own hyperspectral images on board and filtered out the cloud-covered ones before sending data down.",
   "marks": [
    {
     "stage": "L",
     "evidence": 3,
     "basis": [
      "c3"
     ]
    },
    {
     "stage": "R",
     "evidence": 1,
     "basis": [
      "c1"
     ]
    }
   ],
   "claims": [
    {
     "id": "c1",
     "kind": "quote",
     "text": "The satellite successfully implements AI-driven image processing in orbit, achieving 3 TOPS processing speed while reducing data transmission through intelligent filtering of cloud-covered images.",
     "speaker": "KP Labs",
     "relation": "operator",
     "source": "kplabs-missions",
     "url": "https://www.kplabs.space/missions",
     "source_title": "Missions",
     "publisher": "KP Labs",
     "source_kind": "product-page",
     "published": "undated",
     "published_basis": "no date on the page",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in part: about the first 1,500 of 4,500 characters"
    },
    {
     "id": "c2",
     "kind": "quote",
     "text": "a 6U satellite equipped with a 192-band hyperspectral sensor and Leopard Data Processing Unit, launched in November 2023",
     "speaker": "KP Labs",
     "relation": "operator",
     "source": "kplabs-missions",
     "url": "https://www.kplabs.space/missions",
     "source_title": "Missions",
     "publisher": "KP Labs",
     "source_kind": "product-page",
     "published": "undated",
     "published_basis": "no date on the page",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in part: about the first 1,500 of 4,500 characters"
    },
    {
     "id": "c3",
     "kind": "data",
     "fields": {
      "OBJECT_NAME": "INTUITION-1",
      "LAUNCH_DATE": "2023-11-11",
      "DECAY_DATE": "2026-03-23"
     },
     "speaker": "CelesTrak satellite catalogue",
     "relation": "catalogue",
     "source": "celestrak-58565",
     "url": "https://celestrak.org/satcat/records.php?CATNR=58565&FORMAT=json",
     "source_title": "Satellite catalogue record 58565",
     "publisher": "CelesTrak",
     "source_kind": "catalogue",
     "published": "undated",
     "published_basis": "a live catalogue record; the access date is the date that matters",
     "accessed": "2026-10-04",
     "copy_route": "json",
     "read": "in full"
    }
   ],
   "conflicts": [
    {
     "summary": "The operator's page speaks of the mission in the present tense. The catalogue records the satellite as decayed on 23 March 2026.",
     "status": "the page is out of date"
    }
   ],
   "notes": "One sentence on a company's portfolio page. No paper was looked for.",
   "changed": "Launch confirmed from the catalogue, which also shows the satellite has since re-entered.",
   "headline": "L3 R1"
  },
  {
   "id": "dynamic-targeting",
   "entry": 10,
   "name": "Dynamic Targeting on CogniSAT-6",
   "group": "ran",
   "class": "perception",
   "location": "Earth orbit",
   "operator": "Open Cosmos (the satellite)",
   "developer": "NASA Jet Propulsion Laboratory (the software); Ubotica (the payload)",
   "host": "CogniSAT-6",
   "catalogue": null,
   "dates": {
    "launched": "2024-03",
    "ran": "mid-July 2025",
    "reported": "2025-07-24",
    "ended": null
   },
   "hardware": "\"a commercially available AI processor\", not named",
   "software": "An algorithm trained to identify clouds, with planning software",
   "did": "Looked ahead along its orbit, told cloud from clear sky and decided where to point its instrument, with no human involved.",
   "marks": [
    {
     "stage": "L",
     "evidence": 1,
     "basis": [
      "c4"
     ]
    },
    {
     "stage": "R",
     "evidence": 1,
     "basis": [
      "c1",
      "c2"
     ]
    }
   ],
   "claims": [
    {
     "id": "c1",
     "kind": "quote",
     "text": "The technology enabled an Earth-observing satellite for the first time to look ahead along its orbital path, rapidly process and analyze imagery with onboard AI, and determine where to point an instrument.",
     "speaker": "NASA JPL",
     "relation": "developer",
     "source": "jpl-dynamic-targeting",
     "url": "https://www.jpl.nasa.gov/news/how-nasa-is-testing-ai-to-make-earth-observing-satellites-smarter/",
     "source_title": "How NASA Is Testing AI to Make Earth-Observing Satellites Smarter",
     "publisher": "NASA Jet Propulsion Laboratory",
     "source_kind": "press-release",
     "published": "2025-07-24",
     "published_basis": "page metadata",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in full"
    },
    {
     "id": "c2",
     "kind": "quote",
     "text": "The first of a series of flight tests occurred aboard a commercial satellite in mid-July.",
     "speaker": "NASA JPL",
     "relation": "developer",
     "source": "jpl-dynamic-targeting",
     "url": "https://www.jpl.nasa.gov/news/how-nasa-is-testing-ai-to-make-earth-observing-satellites-smarter/",
     "source_title": "How NASA Is Testing AI to Make Earth-Observing Satellites Smarter",
     "publisher": "NASA Jet Propulsion Laboratory",
     "source_kind": "press-release",
     "published": "2025-07-24",
     "published_basis": "page metadata",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in full"
    },
    {
     "id": "c3",
     "kind": "quote",
     "text": "hosts a payload designed and developed by Ubotica featuring a commercially available AI processor",
     "speaker": "NASA JPL",
     "relation": "developer",
     "source": "jpl-dynamic-targeting",
     "url": "https://www.jpl.nasa.gov/news/how-nasa-is-testing-ai-to-make-earth-observing-satellites-smarter/",
     "source_title": "How NASA Is Testing AI to Make Earth-Observing Satellites Smarter",
     "publisher": "NASA Jet Propulsion Laboratory",
     "source_kind": "press-release",
     "published": "2025-07-24",
     "published_basis": "page metadata",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in full"
    },
    {
     "id": "c4",
     "kind": "quote",
     "text": "CogniSAT-6, a briefcase-size CubeSat that launched in March 2024",
     "speaker": "NASA JPL",
     "relation": "developer",
     "source": "jpl-dynamic-targeting",
     "url": "https://www.jpl.nasa.gov/news/how-nasa-is-testing-ai-to-make-earth-observing-satellites-smarter/",
     "source_title": "How NASA Is Testing AI to Make Earth-Observing Satellites Smarter",
     "publisher": "NASA Jet Propulsion Laboratory",
     "source_kind": "press-release",
     "published": "2025-07-24",
     "published_basis": "page metadata",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in full"
    }
   ],
   "conflicts": [],
   "notes": "The statement comes from the maker of the software, about a flight on another company's satellite. Neither Open Cosmos nor Ubotica was read. No catalogue object was found under the name tried.",
   "changed": null,
   "headline": "L1 R1"
  },
  {
   "id": "wukong-tiangong",
   "entry": null,
   "name": "Wukong AI on the Tiangong space station",
   "group": "ran",
   "class": "language",
   "location": "space station",
   "operator": "China's astronaut research and training centre",
   "developer": "not named in what was read",
   "host": "Tiangong space station",
   "catalogue": null,
   "dates": {
    "launched": "2025-07-15",
    "ran": "from mid-July 2025",
    "reported": "2025-08-16",
    "ended": null
   },
   "hardware": "not stated in what was read",
   "software": "A large language model built on an open-source Chinese model, not named",
   "did": "Answered the crew's questions, including the work schedule before a spacewalk.",
   "marks": [
    {
     "stage": "L",
     "evidence": 1,
     "basis": [
      "c1"
     ]
    },
    {
     "stage": "R",
     "evidence": 1,
     "basis": [
      "c2",
      "c4"
     ]
    }
   ],
   "claims": [
    {
     "id": "c1",
     "kind": "quote",
     "text": "the large-scale artificial intelligence model called Wukong AI was delivered to the Chinese space station by the Tianzhou-9 cargo craft on July 15",
     "speaker": "Xinhua, relaying the astronaut training centre",
     "relation": "operator-via-state-media",
     "source": "xinhua-wukong",
     "url": "https://english.news.cn/20250816/f4663eceab5946918607e33f824ebb19/c.html",
     "source_title": "Tiangong space station taikonauts gain AI support",
     "publisher": "Xinhua",
     "source_kind": "news",
     "published": "2025-08-16",
     "published_basis": "page metadata",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in part: five sentences of 3,100 characters"
    },
    {
     "id": "c2",
     "kind": "quote",
     "text": "Wukong AI has been operating stably in orbit for one month",
     "speaker": "Xinhua, relaying the astronaut training centre",
     "relation": "operator-via-state-media",
     "source": "xinhua-wukong",
     "url": "https://english.news.cn/20250816/f4663eceab5946918607e33f824ebb19/c.html",
     "source_title": "Tiangong space station taikonauts gain AI support",
     "publisher": "Xinhua",
     "source_kind": "news",
     "published": "2025-08-16",
     "published_basis": "page metadata",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in part: five sentences of 3,100 characters"
    },
    {
     "id": "c3",
     "kind": "quote",
     "text": "This marks the first time that China's space station has applied and verified large-scale AI model technology.",
     "speaker": "Xinhua, relaying the astronaut training centre",
     "relation": "operator-via-state-media",
     "source": "xinhua-wukong",
     "url": "https://english.news.cn/20250816/f4663eceab5946918607e33f824ebb19/c.html",
     "source_title": "Tiangong space station taikonauts gain AI support",
     "publisher": "Xinhua",
     "source_kind": "news",
     "published": "2025-08-16",
     "published_basis": "page metadata",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in part: five sentences of 3,100 characters"
    },
    {
     "id": "c4",
     "kind": "quote",
     "text": "Commander Chen Dong and taikonaut Wang Jie asked Wukong AI for the work schedule a day before conducting their third spacewalk.",
     "speaker": "Xinhua, relaying the astronaut training centre",
     "relation": "operator-via-state-media",
     "source": "xinhua-wukong",
     "url": "https://english.news.cn/20250816/f4663eceab5946918607e33f824ebb19/c.html",
     "source_title": "Tiangong space station taikonauts gain AI support",
     "publisher": "Xinhua",
     "source_kind": "news",
     "published": "2025-08-16",
     "published_basis": "page metadata",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in part: five sentences of 3,100 characters"
    },
    {
     "id": "c5",
     "kind": "quote",
     "text": "the chatbot was introduced on the space station in mid-July, and has already completed its first mission: supporting three taikonauts during a spacewalk",
     "speaker": "Wired",
     "relation": "news",
     "source": "wired-wukong",
     "url": "https://www.wired.com/story/meet-wukong-chinas-first-ai-chatbot-on-a-space-station/",
     "source_title": "Meet Wukong, the AI Chatbot China Has Installed on Its Space Station",
     "publisher": "Wired",
     "source_kind": "news",
     "published": "2025-08-21",
     "published_basis": "page metadata",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in part: three sentences of 3,400 characters"
    },
    {
     "id": "c6",
     "kind": "quote",
     "text": "Built on a home-grown open-source AI model, Wukong AI is designed to meet the requirements of manned space missions.",
     "speaker": "Xinhua, relaying the astronaut training centre",
     "relation": "operator-via-state-media",
     "source": "xinhua-wukong",
     "url": "https://english.news.cn/20250816/f4663eceab5946918607e33f824ebb19/c.html",
     "source_title": "Tiangong space station taikonauts gain AI support",
     "publisher": "Xinhua",
     "source_kind": "news",
     "published": "2025-08-16",
     "published_basis": "page metadata",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in part: five sentences of 3,100 characters"
    }
   ],
   "conflicts": [],
   "notes": "What was read does not say how much of the system runs on the station and how much on the ground. The report says the model was delivered by cargo craft and has operated in orbit, and it is recorded on that. Chinese-language sources were not read.",
   "changed": "Added on 4 October 2026, in the second round of reading.",
   "headline": "L1 R1"
  },
  {
   "id": "starcloud-1",
   "entry": 11,
   "name": "Starcloud-1",
   "group": "ran",
   "class": "language",
   "location": "Earth orbit",
   "operator": "Starcloud",
   "developer": "Starcloud",
   "host": "Starcloud-1",
   "catalogue": {
    "number": 66303,
    "name": "STARCLOUD-1",
    "launched": "2025-11-02",
    "decayed": null,
    "match": "by name, the reviewer's judgement"
   },
   "dates": {
    "launched": "2025-11-02",
    "ran": "December 2025",
    "reported": "2025-12-10",
    "ended": null
   },
   "hardware": "NVIDIA H100",
   "software": "NanoGPT, trained in orbit; then a Google model, which sources name differently",
   "did": "Trained a small language model in orbit and ran a larger one.",
   "marks": [
    {
     "stage": "L",
     "evidence": 3,
     "basis": [
      "c6"
     ]
    },
    {
     "stage": "R",
     "evidence": 1,
     "basis": [
      "c2"
     ]
    }
   ],
   "claims": [
    {
     "id": "c1",
     "kind": "quote",
     "text": "launched in November 2025 with the first Nvidia H100 GPU in space",
     "speaker": "Starcloud",
     "relation": "operator",
     "source": "starcloud-1-page",
     "url": "https://www.starcloud.com/starcloud-1",
     "source_title": "Starcloud-1",
     "publisher": "Starcloud",
     "source_kind": "product-page",
     "published": "undated",
     "published_basis": "no date in the text; page metadata says 2025-12-10",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in full as extracted: 990 characters of a much larger page, so parts drawn by script may be missing"
    },
    {
     "id": "c2",
     "kind": "quote",
     "text": "became the first satellite to run a version of Gemini in space and the first spacecraft to train an LLM",
     "speaker": "Starcloud",
     "relation": "operator",
     "source": "starcloud-1-page",
     "url": "https://www.starcloud.com/starcloud-1",
     "source_title": "Starcloud-1",
     "publisher": "Starcloud",
     "source_kind": "product-page",
     "published": "undated",
     "published_basis": "no date in the text; page metadata says 2025-12-10",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in full as extracted: 990 characters of a much larger page, so parts drawn by script may be missing"
    },
    {
     "id": "c3",
     "kind": "quote",
     "text": "running and querying responses from Gemma, an open large language model from Google",
     "speaker": "CNBC",
     "relation": "news",
     "source": "cnbc-starcloud",
     "url": "https://www.cnbc.com/2025/12/10/nvidia-backed-starcloud-trains-first-ai-model-in-space-orbital-data-centers.html",
     "source_title": "'Greetings, earthlings': Nvidia-backed Starcloud trains first AI model in space as orbital data center race heats up",
     "publisher": "CNBC",
     "source_kind": "news",
     "published": "2025-12-10",
     "published_basis": "page metadata",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in full"
    },
    {
     "id": "c4",
     "kind": "quote",
     "text": "trains first AI model in space",
     "speaker": "CNBC, in its headline",
     "relation": "news",
     "source": "cnbc-starcloud",
     "url": "https://www.cnbc.com/2025/12/10/nvidia-backed-starcloud-trains-first-ai-model-in-space-orbital-data-centers.html",
     "source_title": "'Greetings, earthlings': Nvidia-backed Starcloud trains first AI model in space as orbital data center race heats up",
     "publisher": "CNBC",
     "source_kind": "news",
     "published": "2025-12-10",
     "published_basis": "page metadata",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in full"
    },
    {
     "id": "c5",
     "kind": "quote",
     "text": "the first time in history that an LLM has been run on a high-powered Nvidia GPU in outer space",
     "speaker": "CNBC",
     "relation": "news",
     "source": "cnbc-starcloud",
     "url": "https://www.cnbc.com/2025/12/10/nvidia-backed-starcloud-trains-first-ai-model-in-space-orbital-data-centers.html",
     "source_title": "'Greetings, earthlings': Nvidia-backed Starcloud trains first AI model in space as orbital data center race heats up",
     "publisher": "CNBC",
     "source_kind": "news",
     "published": "2025-12-10",
     "published_basis": "page metadata",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in full"
    },
    {
     "id": "c6",
     "kind": "data",
     "fields": {
      "OBJECT_NAME": "STARCLOUD-1",
      "LAUNCH_DATE": "2025-11-02"
     },
     "speaker": "CelesTrak satellite catalogue",
     "relation": "catalogue",
     "source": "celestrak-66303",
     "url": "https://celestrak.org/satcat/records.php?CATNR=66303&FORMAT=json",
     "source_title": "Satellite catalogue record 66303",
     "publisher": "CelesTrak",
     "source_kind": "catalogue",
     "published": "undated",
     "published_basis": "a live catalogue record; the access date is the date that matters",
     "accessed": "2026-10-04",
     "copy_route": "json",
     "read": "in full"
    },
    {
     "id": "c7",
     "kind": "quote",
     "text": "Starcloud-1 produced better results with a pre-trained version of Gemma, an open-source AI model from Google.",
     "speaker": "GeekWire",
     "relation": "news",
     "source": "geekwire-starcloud",
     "url": "https://www.geekwire.com/2025/starcloud-power-training-ai-space/",
     "source_title": "Starcloud plans its next power plays after training first AI model in space",
     "publisher": "GeekWire",
     "source_kind": "news",
     "published": "2025-12-22",
     "published_basis": "page metadata",
     "accessed": "2026-10-04",
     "copy_route": "wayback",
     "read": "in part: three sentences of 8,100 characters, from an Internet Archive copy"
    },
    {
     "id": "c8",
     "kind": "quote",
     "text": "Using the Nvidia chip, Starcloud trained a large language model called NanoGPT, a feat that hadn't been done in space before.",
     "speaker": "GeekWire",
     "relation": "news",
     "source": "geekwire-starcloud",
     "url": "https://www.geekwire.com/2025/starcloud-power-training-ai-space/",
     "source_title": "Starcloud plans its next power plays after training first AI model in space",
     "publisher": "GeekWire",
     "source_kind": "news",
     "published": "2025-12-22",
     "published_basis": "page metadata",
     "accessed": "2026-10-04",
     "copy_route": "wayback",
     "read": "in part: three sentences of 8,100 characters, from an Internet Archive copy"
    }
   ],
   "conflicts": [
    {
     "summary": "Which Google model ran. Starcloud's own page says \"a version of Gemini\". Two news reports, CNBC's and GeekWire's, say Gemma.",
     "status": "unresolved"
    }
   ],
   "notes": "The news report quotes Google DeepMind and NVIDIA in support. Both have a stake, as maker of the model and maker of the chip, so neither counts as confirmation. No logs, data or paper were found.",
   "changed": "A second news report was added on 4 October 2026. It also names the model as Gemma. The marks are unchanged.",
   "headline": "L3 R1"
  },
  {
   "id": "three-body",
   "entry": 12,
   "name": "Three-Body Computing Constellation",
   "group": "ran",
   "class": "perception",
   "location": "Earth orbit",
   "operator": "Zhejiang Lab, with partners",
   "developer": "Zhejiang Lab",
   "host": "The constellation's first twelve satellites",
   "catalogue": null,
   "dates": {
    "launched": "2025-05",
    "ran": "November 2025 for the survey reported",
    "reported": "2026-02-13",
    "ended": null
   },
   "hardware": "not named",
   "software": "Ten models, including an 8-billion-parameter remote sensing model and an 8-billion-parameter astronomical model",
   "did": "Surveyed infrastructure from orbit over an area in north-west China, and classified gamma-ray bursts on two satellites.",
   "marks": [
    {
     "stage": "L",
     "evidence": 1,
     "basis": [
      "c1"
     ]
    },
    {
     "stage": "R",
     "evidence": 1,
     "basis": [
      "c2",
      "c3",
      "c4",
      "c5"
     ]
    }
   ],
   "claims": [
    {
     "id": "c1",
     "kind": "quote",
     "text": "China placed 12 satellites, the first group of the space computing constellation called \"Three-Body Computing Constellation,\" into orbit in May 2025.",
     "speaker": "Xinhua, relaying Zhejiang Lab",
     "relation": "operator-via-state-media",
     "source": "xinhua-three-body",
     "url": "https://english.news.cn/20260213/f697fc260d66410398395307dd27443b/c.html",
     "source_title": "China demonstrates AI computing power in outer space with satellite network breakthrough",
     "publisher": "Xinhua",
     "source_kind": "news",
     "published": "2026-02-13",
     "published_basis": "page metadata",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in full"
    },
    {
     "id": "c2",
     "kind": "quote",
     "text": "8-billion-parameter remote sensing model",
     "speaker": "Xinhua, relaying Zhejiang Lab",
     "relation": "operator-via-state-media",
     "source": "xinhua-three-body",
     "url": "https://english.news.cn/20260213/f697fc260d66410398395307dd27443b/c.html",
     "source_title": "China demonstrates AI computing power in outer space with satellite network breakthrough",
     "publisher": "Xinhua",
     "source_kind": "news",
     "published": "2026-02-13",
     "published_basis": "page metadata",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in full"
    },
    {
     "id": "c3",
     "kind": "quote",
     "text": "with 99 percent accuracy",
     "speaker": "Xinhua, relaying Zhejiang Lab",
     "relation": "operator-via-state-media",
     "source": "xinhua-three-body",
     "url": "https://english.news.cn/20260213/f697fc260d66410398395307dd27443b/c.html",
     "source_title": "China demonstrates AI computing power in outer space with satellite network breakthrough",
     "publisher": "Xinhua",
     "source_kind": "news",
     "published": "2026-02-13",
     "published_basis": "page metadata",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in full"
    },
    {
     "id": "c4",
     "kind": "quote",
     "text": "189 square kilometers in northwest China",
     "speaker": "Xinhua, relaying Zhejiang Lab",
     "relation": "operator-via-state-media",
     "source": "xinhua-three-body",
     "url": "https://english.news.cn/20260213/f697fc260d66410398395307dd27443b/c.html",
     "source_title": "China demonstrates AI computing power in outer space with satellite network breakthrough",
     "publisher": "Xinhua",
     "source_kind": "news",
     "published": "2026-02-13",
     "published_basis": "page metadata",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in full"
    },
    {
     "id": "c5",
     "kind": "quote",
     "text": "10 AI models",
     "speaker": "Xinhua, relaying Zhejiang Lab",
     "relation": "operator-via-state-media",
     "source": "xinhua-three-body",
     "url": "https://english.news.cn/20260213/f697fc260d66410398395307dd27443b/c.html",
     "source_title": "China demonstrates AI computing power in outer space with satellite network breakthrough",
     "publisher": "Xinhua",
     "source_kind": "news",
     "published": "2026-02-13",
     "published_basis": "page metadata",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in full"
    },
    {
     "id": "c6",
     "kind": "quote",
     "text": "China launched 12 satellites early Wednesday for an on-orbit computing project led by startup ADA Space and Zhejiang Lab.",
     "speaker": "SpaceNews",
     "relation": "news",
     "source": "spacenews-three-body",
     "url": "https://spacenews.com/china-launches-first-of-2800-satellites-for-ai-space-computing-constellation/",
     "source_title": "China launches first of 2,800 satellites for AI space computing constellation",
     "publisher": "SpaceNews",
     "source_kind": "news",
     "published": "2025-05-14",
     "published_basis": "page metadata",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in part: the first sentence; the rest is behind a registration wall"
    }
   ],
   "conflicts": [
    {
     "summary": "A third-party tracker says the constellation is \"reported to have run Alibaba's Qwen3\". Its link goes to a news index page, and no source for the statement was found.",
     "status": "unsupported; not recorded as a claim"
    }
   ],
   "notes": "The largest model reported run in space in this register. The account of what ran comes from one English-language report by a state news agency; a trade news report, of which one sentence could be read, reports only the launch. Chinese-language sources were not read. A catalogue object (63982, SCS-01 A, launched 14 May 2025) may belong to this constellation; that match is another tracker's and was not checked, so it is not used.",
   "changed": "A news report of the launch was added on 4 October 2026. The marks are unchanged.",
   "headline": "L1 R1"
  },
  {
   "id": "navi-orbital",
   "entry": 13,
   "name": "NAVI-Orbital on YAM-9",
   "group": "ran",
   "class": "language",
   "location": "Earth orbit",
   "operator": "Loft Orbital (the satellite)",
   "developer": "The paper's author",
   "host": "YAM-9",
   "catalogue": {
    "number": 66694,
    "name": "YAM-9",
    "launched": "2025-11-28",
    "decayed": null,
    "match": "by name, the reviewer's judgement"
   },
   "dates": {
    "launched": "2025-11-28",
    "ran": "2026-04-16",
    "reported": "2026-06-05",
    "ended": null
   },
   "hardware": "\"a heterogeneous compute cluster\"; the flight processor is not named",
   "software": "Gemma 3 4B, a vision-language model",
   "did": "Classified the scenes the satellite captured, described them in text and answered an operator's follow-up questions, all on board.",
   "marks": [
    {
     "stage": "L",
     "evidence": 3,
     "basis": [
      "c4"
     ]
    },
    {
     "stage": "R",
     "evidence": 2,
     "basis": [
      "c1",
      "c2",
      "c3"
     ]
    }
   ],
   "claims": [
    {
     "id": "c1",
     "kind": "quote",
     "text": "On April 16, 2026, NAVI-Orbital achieved what is, to the authors' knowledge, the first in-orbit demonstration of a vision-language model performing autonomous multi-modal inference entirely onboard.",
     "speaker": "The paper's author",
     "relation": "experimenter",
     "source": "arxiv-navi-abs",
     "url": "https://arxiv.org/abs/2606.18271",
     "source_title": "NAVI-Orbital: First In-Orbit Demonstration of a Zero-Shot Vision-Language Model for Autonomous Earth Observation",
     "publisher": "arXiv (preprint by the system's authors)",
     "source_kind": "preprint",
     "published": "2026-06-05",
     "published_basis": "submission date on the page",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "abstract page in full"
    },
    {
     "id": "c2",
     "kind": "quote",
     "text": "Loft Orbital's YAM-9 satellite, a multi-mission platform flying a heterogeneous compute cluster",
     "speaker": "The paper's author",
     "relation": "experimenter",
     "source": "arxiv-navi-html",
     "url": "https://arxiv.org/html/2606.18271v1",
     "source_title": "NAVI-Orbital: First In-Orbit Demonstration of a Zero-Shot Vision-Language Model for Autonomous Earth Observation (full text)",
     "publisher": "arXiv (preprint by the system's authors)",
     "source_kind": "preprint",
     "published": "2026-06-05",
     "published_basis": "submission date on the abstract page",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in part: about 15 excerpts of 74,500 characters"
    },
    {
     "id": "c3",
     "kind": "quote",
     "text": "Gemma 3 4B",
     "speaker": "The paper's author",
     "relation": "experimenter",
     "source": "arxiv-navi-html",
     "url": "https://arxiv.org/html/2606.18271v1",
     "source_title": "NAVI-Orbital: First In-Orbit Demonstration of a Zero-Shot Vision-Language Model for Autonomous Earth Observation (full text)",
     "publisher": "arXiv (preprint by the system's authors)",
     "source_kind": "preprint",
     "published": "2026-06-05",
     "published_basis": "submission date on the abstract page",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in part: about 15 excerpts of 74,500 characters"
    },
    {
     "id": "c4",
     "kind": "data",
     "fields": {
      "OBJECT_NAME": "YAM-9",
      "LAUNCH_DATE": "2025-11-28"
     },
     "speaker": "CelesTrak satellite catalogue",
     "relation": "catalogue",
     "source": "celestrak-66694",
     "url": "https://celestrak.org/satcat/records.php?CATNR=66694&FORMAT=json",
     "source_title": "Satellite catalogue record 66694",
     "publisher": "CelesTrak",
     "source_kind": "catalogue",
     "published": "undated",
     "published_basis": "a live catalogue record; the access date is the date that matters",
     "accessed": "2026-10-04",
     "copy_route": "json",
     "read": "in full"
    }
   ],
   "conflicts": [],
   "notes": "A preprint, not peer reviewed as far as its page shows. The accuracy it reports (88.16%) is for a benchmark on the ground; no in-orbit accuracy was found in the parts read.",
   "changed": "Launch of the host satellite confirmed from the catalogue.",
   "headline": "L3 R2"
  },
  {
   "id": "prithvi-imagin-e",
   "entry": 14,
   "name": "Prithvi on the IMAGIN-e payload, International Space Station",
   "group": "ran",
   "class": "perception",
   "location": "space station",
   "operator": "Thales Alenia Space (the payload)",
   "developer": "Adelaide University, with ESA Φ-lab, Thales Alenia Space and the SmartSat CRC",
   "host": "IMAGIN-e, on the station's Bartolomeo platform",
   "catalogue": null,
   "dates": {
    "launched": "2024-03-21",
    "ran": "before 1 December 2025; no date given",
    "reported": "2025-12-01",
    "ended": null
   },
   "hardware": "an ARM-based compute module with no hardware accelerator",
   "software": "A compressed variant of the NASA and IBM Prithvi geospatial foundation model",
   "did": "Ran inference on a prepared test set in orbit, under command from the ground, and returned outputs matching those obtained on Earth.",
   "marks": [
    {
     "stage": "L",
     "evidence": 2,
     "basis": [
      "c6"
     ]
    },
    {
     "stage": "R",
     "evidence": 2,
     "basis": [
      "c1",
      "c2"
     ]
    }
   ],
   "claims": [
    {
     "id": "c1",
     "kind": "quote",
     "text": "We further demonstrate reliable on-orbit inference with the IMAGIN-e payload aboard the International Space Station.",
     "speaker": "Du et al., the research team",
     "relation": "experimenter",
     "source": "arxiv-prithvi-abs",
     "url": "https://arxiv.org/abs/2512.01181",
     "source_title": "First On-Orbit Demonstration of a Geospatial Foundation Model",
     "publisher": "arXiv (preprint by the demonstration's team)",
     "source_kind": "preprint",
     "published": "2025-12-01",
     "published_basis": "submission date on the page",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "abstract page in full"
    },
    {
     "id": "c2",
     "kind": "quote",
     "text": "Inference was executed onboard the compute module under ground command, with intermediate outputs and runtime logs downlinked for comparison with on-ground baselines.",
     "speaker": "Du et al., the research team",
     "relation": "experimenter",
     "source": "arxiv-prithvi-html",
     "url": "https://arxiv.org/html/2512.01181v1",
     "source_title": "First On-Orbit Demonstration of a Geospatial Foundation Model (full text)",
     "publisher": "arXiv (preprint by the demonstration's team)",
     "source_kind": "preprint",
     "published": "2025-12-01",
     "published_basis": "submission date on the abstract page",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in part: about 25 excerpts of 78,000 characters"
    },
    {
     "id": "c3",
     "kind": "quote",
     "text": "direct on-orbit testing was not possible due to a technical anomaly affecting the EO payload interface",
     "speaker": "Du et al., the research team",
     "relation": "experimenter",
     "source": "arxiv-prithvi-html",
     "url": "https://arxiv.org/html/2512.01181v1",
     "source_title": "First On-Orbit Demonstration of a Geospatial Foundation Model (full text)",
     "publisher": "arXiv (preprint by the demonstration's team)",
     "source_kind": "preprint",
     "published": "2025-12-01",
     "published_basis": "submission date on the abstract page",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in part: about 25 excerpts of 78,000 characters"
    },
    {
     "id": "c4",
     "kind": "quote",
     "text": "this was not possible due to a power fault in the imaging system",
     "speaker": "Du et al., the research team",
     "relation": "experimenter",
     "source": "arxiv-prithvi-html",
     "url": "https://arxiv.org/html/2512.01181v1",
     "source_title": "First On-Orbit Demonstration of a Geospatial Foundation Model (full text)",
     "publisher": "arXiv (preprint by the demonstration's team)",
     "source_kind": "preprint",
     "published": "2025-12-01",
     "published_basis": "submission date on the abstract page",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in part: about 25 excerpts of 78,000 characters"
    },
    {
     "id": "c5",
     "kind": "quote",
     "text": "has successfully uploaded and demonstrated NASA and IBM's open-source Prithvi Geospatial artificial intelligence (AI) foundation model aboard two in-orbit platforms, making it the first geospatial foundation model to be deployed in orbit",
     "speaker": "NASA, which made the model with IBM",
     "relation": "model-maker",
     "source": "nasa-prithvi-orbit",
     "url": "https://science.nasa.gov/science-research/ai-foundation-model-in-orbit/",
     "source_title": "NASA's Prithvi Becomes First AI Geospatial Foundation Model In Orbit",
     "publisher": "NASA Science",
     "source_kind": "press-release",
     "published": "2026-05-07",
     "published_basis": "page metadata; updated 13 May 2026",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in full"
    },
    {
     "id": "c6",
     "kind": "quote",
     "text": "The IMAGIN-e payload was launched on 21 March 2024 aboard a SpaceX cargo resupply mission",
     "speaker": "Du et al., the research team",
     "relation": "experimenter",
     "source": "arxiv-prithvi-html",
     "url": "https://arxiv.org/html/2512.01181v1",
     "source_title": "First On-Orbit Demonstration of a Geospatial Foundation Model (full text)",
     "publisher": "arXiv (preprint by the demonstration's team)",
     "source_kind": "preprint",
     "published": "2025-12-01",
     "published_basis": "submission date on the abstract page",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in part: about 25 excerpts of 78,000 characters"
    }
   ],
   "conflicts": [
    {
     "summary": "How many platforms. NASA's article of May 2026 says the model was demonstrated aboard two in-orbit platforms, the Kanyini satellite and IMAGIN-e. The team's own preprint of December 2025 says direct testing in orbit on Kanyini was not possible, and that the model ran there on an engineering model on the ground. Either the Kanyini run happened later and is reported nowhere read here, or the article counts ground hardware as an orbiting platform.",
     "status": "unresolved; only the IMAGIN-e run is recorded"
    },
    {
     "summary": "The run in orbit used a prepared test set, not images newly taken: the preprint says a power fault in the imaging system prevented that.",
     "status": "noted"
    }
   ],
   "notes": "The preprint thanks Thales Alenia Space for evidence of the demonstration. Thales Alenia Space runs the payload and is among the authors' institutions, so this is not counted as independent.",
   "changed": "Run raised from R1 to R2 after the preprint was read, and narrowed from two platforms to one.",
   "headline": "L2 R2"
  },
  {
   "id": "moi-td",
   "entry": null,
   "name": "MOI-TD",
   "group": "ran",
   "class": "compute",
   "location": "Earth orbit",
   "operator": "TakeMe2Space",
   "developer": "TakeMe2Space",
   "host": "MOI-TD",
   "catalogue": null,
   "dates": {
    "launched": "2024-12-30",
    "ran": "not dated on the page read",
    "reported": null,
    "ended": null
   },
   "hardware": "not stated on the page read",
   "software": "not named",
   "did": "Ran experiments in low Earth orbit that its operator says included AI inference; what was inferred is not stated.",
   "marks": [
    {
     "stage": "R",
     "evidence": 1,
     "basis": [
      "c1"
     ]
    }
   ],
   "claims": [
    {
     "id": "c1",
     "kind": "quote",
     "text": "India's first AI-lab in space. 20+ experiments executed in LEO — on-orbit AI inference, sensor fusion, and high-speed data handling all validated.",
     "speaker": "TakeMe2Space",
     "relation": "operator",
     "source": "tm2-missions",
     "url": "https://www.tm2.space/missions",
     "source_title": "Orbital Computing and Satellite Missions",
     "publisher": "TakeMe2Space",
     "source_kind": "product-page",
     "published": "undated",
     "published_basis": "no date of writing on the page",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in full as extracted: 1,000 characters"
    }
   ],
   "conflicts": [],
   "notes": "The thinnest record with a run mark: one sentence on the operator's missions page, with no model, hardware or result. The page gives the launch as 30 December 2024 from Sriharikota.",
   "changed": "Added on 4 October 2026. It was not among the first twenty entries.",
   "headline": "R1"
  },
  {
   "id": "pelican-4",
   "entry": 20,
   "name": "Planet's Pelican-4: object detection on board",
   "group": "ran",
   "class": "perception",
   "location": "Earth orbit",
   "operator": "Planet Labs PBC",
   "developer": "Planet Labs PBC",
   "host": "Pelican-4",
   "catalogue": {
    "number": 65316,
    "name": "PELICAN-4",
    "launched": "2025-08-26",
    "decayed": null,
    "match": "by name, the reviewer's judgement"
   },
   "dates": {
    "launched": "2025-08-26",
    "ran": "25 March 2026",
    "reported": "2026-04-07",
    "ended": null
   },
   "hardware": "NVIDIA Jetson Orin module",
   "software": "An object-detection model, not named",
   "did": "Took an image of an airport from orbit and ran a model on board to find the aeroplanes in it.",
   "marks": [
    {
     "stage": "L",
     "evidence": 3,
     "basis": [
      "c6"
     ]
    },
    {
     "stage": "R",
     "evidence": 1,
     "basis": [
      "c1",
      "c2",
      "c3"
     ]
    }
   ],
   "claims": [
    {
     "id": "c1",
     "kind": "quote",
     "text": "the successful deployment and execution of AI-driven object detection directly onboard its Pelican-4 satellite",
     "speaker": "Planet Labs PBC",
     "relation": "operator",
     "source": "planet-ai-in-space-release",
     "url": "https://investors.planet.com/news/news-details/2026/Planet-Successfully-Runs-AI-in-Space/default.aspx",
     "source_title": "Planet Successfully Runs AI in Space",
     "publisher": "Planet Labs PBC",
     "source_kind": "press-release",
     "published": "2026-04-07",
     "published_basis": "date printed on the release",
     "accessed": "2026-10-04",
     "copy_route": "wayback",
     "read": "in full: the body of the release, about 3,600 characters, up to its forward-looking statements"
    },
    {
     "id": "c2",
     "kind": "quote",
     "text": "Planet's Pelican-4 captured an image of an airport, then successfully utilized its onboard NVIDIA Jetson Orin module to run an AI model to detect airplanes in moments",
     "speaker": "Planet Labs PBC",
     "relation": "operator",
     "source": "planet-ai-in-space-release",
     "url": "https://investors.planet.com/news/news-details/2026/Planet-Successfully-Runs-AI-in-Space/default.aspx",
     "source_title": "Planet Successfully Runs AI in Space",
     "publisher": "Planet Labs PBC",
     "source_kind": "press-release",
     "published": "2026-04-07",
     "published_basis": "date printed on the release",
     "accessed": "2026-10-04",
     "copy_route": "wayback",
     "read": "in full: the body of the release, about 3,600 characters, up to its forward-looking statements"
    },
    {
     "id": "c3",
     "kind": "quote",
     "text": "This first detection has an 80% detection accuracy on raw imagery.",
     "speaker": "Planet Labs PBC",
     "relation": "operator",
     "source": "planet-ai-in-space-release",
     "url": "https://investors.planet.com/news/news-details/2026/Planet-Successfully-Runs-AI-in-Space/default.aspx",
     "source_title": "Planet Successfully Runs AI in Space",
     "publisher": "Planet Labs PBC",
     "source_kind": "press-release",
     "published": "2026-04-07",
     "published_basis": "date printed on the release",
     "accessed": "2026-10-04",
     "copy_route": "wayback",
     "read": "in full: the body of the release, about 3,600 characters, up to its forward-looking statements"
    },
    {
     "id": "c4",
     "kind": "quote",
     "text": "This represents one of the first times an Earth imaging satellite has moved beyond simple data capture to onboard AI inference and analysis.",
     "speaker": "Planet Labs PBC",
     "relation": "operator",
     "source": "planet-ai-in-space-release",
     "url": "https://investors.planet.com/news/news-details/2026/Planet-Successfully-Runs-AI-in-Space/default.aspx",
     "source_title": "Planet Successfully Runs AI in Space",
     "publisher": "Planet Labs PBC",
     "source_kind": "press-release",
     "published": "2026-04-07",
     "published_basis": "date printed on the release",
     "accessed": "2026-10-04",
     "copy_route": "wayback",
     "read": "in full: the body of the release, about 3,600 characters, up to its forward-looking statements"
    },
    {
     "id": "c5",
     "kind": "quote",
     "text": "was joined by five more in 2025: Pelican-2 in January, Pelican-3 and Pelican-4 in August, and Pelican-5 and Pelican-6 in November",
     "speaker": "Planet Labs PBC",
     "relation": "operator",
     "source": "planet-pelican-2025-review",
     "url": "https://www.planet.com/pulse/pelican-2025-in-review/",
     "source_title": "Pelican 2025 in Review",
     "publisher": "Planet Labs PBC",
     "source_kind": "blog",
     "published": "2026-01-13",
     "published_basis": "page metadata",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in part: the section on launches, about 600 of 4,000 characters"
    },
    {
     "id": "c6",
     "kind": "data",
     "fields": {
      "OBJECT_NAME": "PELICAN-4",
      "LAUNCH_DATE": "2025-08-26"
     },
     "speaker": "CelesTrak satellite catalogue",
     "relation": "catalogue",
     "source": "celestrak-65316",
     "url": "https://celestrak.org/satcat/records.php?CATNR=65316&FORMAT=json",
     "source_title": "Satellite catalogue record 65316 (PELICAN-4)",
     "publisher": "CelesTrak",
     "source_kind": "catalogue",
     "published": "undated",
     "published_basis": "a live catalogue record; dated by when it was read",
     "accessed": "2026-10-04",
     "copy_route": "json",
     "read": "in full"
    },
    {
     "id": "c7",
     "kind": "quote",
     "text": "Planet Labs uses NVIDIA to convert satellite imagery into real-time intelligence in orbit",
     "speaker": "NVIDIA, the chip supplier",
     "relation": "supplier",
     "source": "nvidia-space-computing",
     "url": "https://www.nvidia.com/en-us/edge-computing/space-computing/",
     "source_title": "Space Computing: On-Orbit AI & Accelerated Computing",
     "publisher": "NVIDIA",
     "source_kind": "product-page",
     "published": "undated",
     "published_basis": "no date in the text; page metadata says 2026-09-16",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in part: the first 3,800 of 4,100 characters"
    }
   ],
   "conflicts": [],
   "notes": "The accuracy figure is the operator's, for one image. Its wording of priority is careful: \"one of the first times\", not the first. The release was read from an archive capture of Planet's investor page, because the page itself refused the request.",
   "changed": "Until 4 October 2026 this entry rested on the chip supplier's product page alone, at R0. Planet's own release was then found and read, which makes it an operator's statement of a run: R1. The launch is confirmed from the catalogue.",
   "headline": "L3 R1"
  },
  {
   "id": "moonnet-hakuto-r",
   "entry": null,
   "name": "MoonNet on ispace's HAKUTO-R Mission 1",
   "group": "launched",
   "class": "perception",
   "location": "lunar orbit",
   "operator": "ispace (the lander)",
   "developer": "Mission Control Space Services",
   "host": "HAKUTO-R Mission 1 lander",
   "catalogue": {
    "number": 54696,
    "name": "HAKUTO-R M1",
    "launched": "2022-12-11",
    "decayed": "2023-04-25",
    "match": "by the mission named on the developer's page; the reviewer's judgement"
   },
   "dates": {
    "launched": "2022-12-11",
    "ran": null,
    "reported": null,
    "ended": "2023-04-25"
   },
   "hardware": "a computer on the lander, not named in what was read",
   "software": "MoonNet, a deep learning model to classify lunar terrain",
   "did": "Was carried to lunar orbit to classify a rover's images of the surface. Its maker says it was deployed there; no output is reported in what was read.",
   "marks": [
    {
     "stage": "L",
     "evidence": 3,
     "basis": [
      "c3"
     ]
    }
   ],
   "claims": [
    {
     "id": "c1",
     "kind": "quote",
     "text": "Mission Control was the first organization in the world to deploy a deep-learning Artificial Intelligence (AI) in lunar orbit",
     "speaker": "Mission Control Space Services",
     "relation": "developer",
     "source": "missioncontrol-spacefarer",
     "url": "https://missioncontrolspace.com/products/spacefarer-ai/",
     "source_title": "Spacefarer AI",
     "publisher": "Mission Control Space Services",
     "source_kind": "product-page",
     "published": "undated",
     "published_basis": "no date seen",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in part: three sentences of 3,200 characters"
    },
    {
     "id": "c2",
     "kind": "quote",
     "text": "will host MoonNet, an Artificial Intelligence (AI) application to classify types of lunar geological features",
     "speaker": "Mission Control Space Services",
     "relation": "developer",
     "source": "missioncontrol-lunar-ai",
     "url": "https://missioncontrolspace.com/stories/lunar-ai-demo-and-the-ispace-m1-mission/",
     "source_title": "Lunar AI demo and the ispace M1 mission",
     "publisher": "Mission Control Space Services",
     "source_kind": "blog",
     "published": "undated",
     "published_basis": "no date seen; written before the flight",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in part: seven sentences of 3,600 characters"
    },
    {
     "id": "c3",
     "kind": "data",
     "fields": {
      "OBJECT_NAME": "HAKUTO-R M1",
      "LAUNCH_DATE": "2022-12-11",
      "DECAY_DATE": "2023-04-25"
     },
     "speaker": "CelesTrak satellite catalogue",
     "relation": "catalogue",
     "source": "celestrak-54696",
     "url": "https://celestrak.org/satcat/records.php?CATNR=54696&FORMAT=json",
     "source_title": "Satellite catalogue record 54696 (HAKUTO-R M1)",
     "publisher": "CelesTrak",
     "source_kind": "catalogue",
     "published": "undated",
     "published_basis": "a live catalogue record; dated by when it was read",
     "accessed": "2026-10-04",
     "copy_route": "json",
     "read": "in full"
    }
   ],
   "conflicts": [
    {
     "summary": "Deployed or run. The maker says its model was deployed in lunar orbit. The pages read were written before the flight or say nothing of what the model produced. The catalogue gives the lander's end as 25 April 2023.",
     "status": "unresolved; recorded as launched"
    }
   ],
   "notes": "The model was meant to work on images from a rover after landing. What became of the landing is not in what was read.",
   "changed": "Added on 4 October 2026, in the second round of reading.",
   "headline": "L3"
  },
  {
   "id": "scenic-iss",
   "entry": null,
   "name": "SpaceCube SCENIC on the International Space Station",
   "group": "launched",
   "class": "compute",
   "location": "space station",
   "operator": "NASA Goddard Space Flight Center",
   "developer": "NASA Goddard Space Flight Center",
   "host": "Space Test Program-Houston 9, on the International Space Station",
   "catalogue": null,
   "dates": {
    "launched": "2023-03",
    "ran": null,
    "reported": "2023",
    "ended": null
   },
   "hardware": "SpaceCube processors: FPGAs and custom microchips",
   "software": "none reported",
   "did": "Was sent to the station as a testbed for AI and machine learning. What was read describes the plan and reports no result.",
   "marks": [
    {
     "stage": "L",
     "evidence": 1,
     "basis": [
      "c1"
     ]
    }
   ],
   "claims": [
    {
     "id": "c1",
     "kind": "quote",
     "text": "in March 2023, the SpaceCube Edge Node Intelligent Collaboration (SCENIC) experiment was launched to the International Space Station (ISS) on board the DoD Space Test Program-Houston 9 (STP-H9) mission",
     "speaker": "NASA Earth Science Technology Office",
     "relation": "developer",
     "source": "nasa-esto-spacecube",
     "url": "https://esto.nasa.gov/25years/spacecube/",
     "source_title": "SpaceCube: Powerful Processing in Space",
     "publisher": "NASA Earth Science Technology Office",
     "source_kind": "project-page",
     "published": "undated",
     "published_basis": "no date seen",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in part: three sentences of 2,800 characters"
    },
    {
     "id": "c2",
     "kind": "quote",
     "text": "is a testbed to evaluate artificial intelligence and machine learning on FPGAs and custom microchips",
     "speaker": "NASA Earth Science Technology Office",
     "relation": "developer",
     "source": "nasa-esto-spacecube",
     "url": "https://esto.nasa.gov/25years/spacecube/",
     "source_title": "SpaceCube: Powerful Processing in Space",
     "publisher": "NASA Earth Science Technology Office",
     "source_kind": "project-page",
     "published": "undated",
     "published_basis": "no date seen",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in part: three sentences of 2,800 characters"
    },
    {
     "id": "c3",
     "kind": "quote",
     "text": "will demonstrate inflight, cutting-edge AI applications on multiple space-based devices",
     "speaker": "Geist et al., NASA Goddard, the experiment's team",
     "relation": "mission-team",
     "source": "smallsat-scenic",
     "url": "https://digitalcommons.usu.edu/smallsat/2023/all2023/147/",
     "source_title": "NASA SpaceCube Next-Generation Artificial-Intelligence Computing for STP-H9-SCENIC on ISS",
     "publisher": "Small Satellite Conference 2023 (Geist and co-authors, NASA Goddard, the experiment's team)",
     "source_kind": "abstract",
     "published": "2023",
     "published_basis": "conference year in the address",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in part: four sentences of 5,600 characters"
    }
   ],
   "conflicts": [],
   "notes": "Both sources were written as plans. Results may have been published since and were not looked for.",
   "changed": "Added on 4 October 2026, in the second round of reading.",
   "headline": "L1"
  },
  {
   "id": "yam-6",
   "entry": null,
   "name": "Loft Orbital's YAM-6",
   "group": "launched",
   "class": "compute",
   "location": "Earth orbit",
   "operator": "Loft Orbital",
   "developer": "Loft Orbital",
   "host": "YAM-6",
   "catalogue": {
    "number": 59126,
    "name": "YAM-6",
    "launched": "2024-03-04",
    "decayed": null,
    "match": "by name, the reviewer's judgement"
   },
   "dates": {
    "launched": "2024-03-04",
    "ran": null,
    "reported": "2025-01-14",
    "ended": null
   },
   "hardware": "processors and graphics processors, not named in what was read",
   "software": "customers' own software; none named",
   "did": "Flies as a satellite on which customers run their own software. The operator says it runs their missions daily; no AI workload is named in what was read.",
   "marks": [
    {
     "stage": "L",
     "evidence": 3,
     "basis": [
      "c5"
     ]
    }
   ],
   "claims": [
    {
     "id": "c1",
     "kind": "quote",
     "text": "Launched YAM-6, the world's first satellite dedicated to running AI in space",
     "speaker": "Loft Orbital",
     "relation": "operator",
     "source": "loft-series-c",
     "url": "https://loftorbital.com/announcing-lofts-170m-series-c/",
     "source_title": "Announcing Loft's $170M Series C",
     "publisher": "Loft Orbital",
     "source_kind": "blog",
     "published": "2025-01-14",
     "published_basis": "page metadata",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in part: two sentences of 9,700 characters"
    },
    {
     "id": "c2",
     "kind": "quote",
     "text": "Today, we are running virtual missions on YAM-6 every day for partners and customers such as Helsing, Microsoft, Agenium, NTT, and many others.",
     "speaker": "Loft Orbital",
     "relation": "operator",
     "source": "loft-series-c",
     "url": "https://loftorbital.com/announcing-lofts-170m-series-c/",
     "source_title": "Announcing Loft's $170M Series C",
     "publisher": "Loft Orbital",
     "source_kind": "blog",
     "published": "2025-01-14",
     "published_basis": "page metadata",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in part: two sentences of 9,700 characters"
    },
    {
     "id": "c3",
     "kind": "quote",
     "text": "Launching on Transporter-10, YAM-6 will abstract away the hardware by providing access to Loft-owned sensors and compute nodes that support AI.",
     "speaker": "Loft Orbital",
     "relation": "operator",
     "source": "loft-yam6",
     "url": "https://loftorbital.com/yam-6-the-rise-of-the-virtual-mission/",
     "source_title": "YAM-6: The Rise of the Virtual Mission",
     "publisher": "Loft Orbital",
     "source_kind": "blog",
     "published": "2023-11-21",
     "published_basis": "page metadata",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in part: three sentences of 5,400 characters"
    },
    {
     "id": "c4",
     "kind": "quote",
     "text": "We've developed software that lets our customers deploy their own AI models directly onboard our satellites.",
     "speaker": "Loft Orbital",
     "relation": "operator",
     "source": "loft-ai",
     "url": "https://loftorbital.com/inside-loft-orbital-building-smarter-satellites-with-ai/",
     "source_title": "Inside Loft Orbital: Building smarter satellites with AI",
     "publisher": "Loft Orbital",
     "source_kind": "blog",
     "published": "2025-06-30",
     "published_basis": "page metadata",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in part: four sentences of 8,100 characters"
    },
    {
     "id": "c5",
     "kind": "data",
     "fields": {
      "OBJECT_NAME": "YAM-6",
      "LAUNCH_DATE": "2024-03-04"
     },
     "speaker": "CelesTrak satellite catalogue",
     "relation": "catalogue",
     "source": "celestrak-59126",
     "url": "https://celestrak.org/satcat/records.php?CATNR=59126&FORMAT=json",
     "source_title": "Satellite catalogue record 59126 (YAM-6)",
     "publisher": "CelesTrak",
     "source_kind": "catalogue",
     "published": "undated",
     "published_basis": "a live catalogue record; dated by when it was read",
     "accessed": "2026-10-04",
     "copy_route": "json",
     "read": "in full"
    }
   ],
   "conflicts": [],
   "notes": "Listed as launched, not as run, by a narrow reading: the operator says missions run on it every day and that customers can put AI models on its satellites, but names no model that has run on this one. Its sister satellite YAM-9 has a run on record under another entry.",
   "changed": "Added on 4 October 2026, in the second round of reading.",
   "headline": "L3"
  },
  {
   "id": "lizziesat-3",
   "entry": null,
   "name": "Sidus Space's FeatherEdge computer on LizzieSat-3",
   "group": "launched",
   "class": "compute",
   "location": "Earth orbit",
   "operator": "Sidus Space",
   "developer": "Sidus Space",
   "host": "LizzieSat-3",
   "catalogue": {
    "number": 63233,
    "name": "LIZZIESAT-3 (LS-3)",
    "launched": "2025-03-15",
    "decayed": null,
    "match": "by name, the reviewer's judgement"
   },
   "dates": {
    "launched": "2025-03-15",
    "ran": null,
    "reported": "2025-05-16",
    "ended": null
   },
   "hardware": "FeatherEdge GEN-2, built on an NVIDIA Jetson Orin NX module",
   "software": "none reported",
   "did": "Was reported working in orbit by its operator, which sells its AI processing power. No AI workload is named as run in what was read.",
   "marks": [
    {
     "stage": "L",
     "evidence": 3,
     "basis": [
      "c4"
     ]
    }
   ],
   "claims": [
    {
     "id": "c1",
     "kind": "quote",
     "text": "FeatherEdge™ GEN-2 is fully operational onboard LizzieSat®-3",
     "speaker": "Sidus Space",
     "relation": "operator",
     "source": "sidus-featheredge",
     "url": "https://investors.sidusspace.com/news-events/press-releases/detail/237/sidus-space-announces-successful-on-orbit-operation-of",
     "source_title": "Sidus Space Announces Successful On-Orbit Operation of FeatherEdge Gen-2 Aboard LizzieSat-3",
     "publisher": "Sidus Space",
     "source_kind": "press-release",
     "published": "2025-05-16",
     "published_basis": "page metadata",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in part: about eight sentences of 5,400 characters"
    },
    {
     "id": "c2",
     "kind": "quote",
     "text": "100 trillion operations per second (TOPS) of AI processing power",
     "speaker": "Sidus Space",
     "relation": "operator",
     "source": "sidus-featheredge",
     "url": "https://investors.sidusspace.com/news-events/press-releases/detail/237/sidus-space-announces-successful-on-orbit-operation-of",
     "source_title": "Sidus Space Announces Successful On-Orbit Operation of FeatherEdge Gen-2 Aboard LizzieSat-3",
     "publisher": "Sidus Space",
     "source_kind": "press-release",
     "published": "2025-05-16",
     "published_basis": "page metadata",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in part: about eight sentences of 5,400 characters"
    },
    {
     "id": "c3",
     "kind": "quote",
     "text": "NVIDIA's Jetson Orin™ NX module",
     "speaker": "Sidus Space",
     "relation": "operator",
     "source": "sidus-featheredge",
     "url": "https://investors.sidusspace.com/news-events/press-releases/detail/237/sidus-space-announces-successful-on-orbit-operation-of",
     "source_title": "Sidus Space Announces Successful On-Orbit Operation of FeatherEdge Gen-2 Aboard LizzieSat-3",
     "publisher": "Sidus Space",
     "source_kind": "press-release",
     "published": "2025-05-16",
     "published_basis": "page metadata",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in part: about eight sentences of 5,400 characters"
    },
    {
     "id": "c4",
     "kind": "data",
     "fields": {
      "OBJECT_NAME": "LIZZIESAT-3 (LS-3)",
      "LAUNCH_DATE": "2025-03-15"
     },
     "speaker": "CelesTrak satellite catalogue",
     "relation": "catalogue",
     "source": "celestrak-63233",
     "url": "https://celestrak.org/satcat/records.php?CATNR=63233&FORMAT=json",
     "source_title": "Satellite catalogue record 63233 (LIZZIESAT-3 (LS-3))",
     "publisher": "CelesTrak",
     "source_kind": "catalogue",
     "published": "undated",
     "published_basis": "a live catalogue record; dated by when it was read",
     "accessed": "2026-10-04",
     "copy_route": "json",
     "read": "in full"
    }
   ],
   "conflicts": [],
   "notes": "Only one release was read. The silence recorded here may be this register's.",
   "changed": "Added on 4 October 2026, in the second round of reading.",
   "headline": "L3"
  },
  {
   "id": "axdcu-1",
   "entry": 19,
   "name": "AxDCU-1 on the International Space Station",
   "group": "launched",
   "class": "compute",
   "location": "space station",
   "operator": "Axiom Space",
   "developer": "Axiom Space, with Red Hat",
   "host": "International Space Station",
   "catalogue": null,
   "dates": {
    "launched": "2025-08",
    "ran": null,
    "reported": "2025-09-16",
    "ended": null
   },
   "hardware": "a data processing prototype running Red Hat Device Edge",
   "software": "none reported",
   "did": "Was sent to the station to test cloud computing and, by the plan announced, AI and machine learning. No AI workload is reported in what was read.",
   "marks": [
    {
     "stage": "L",
     "evidence": 1,
     "basis": [
      "c1"
     ]
    }
   ],
   "claims": [
    {
     "id": "c1",
     "kind": "quote",
     "text": "the achievement of launching the Axiom Data Center Unit One (AxDCU-1) to the space station in August",
     "speaker": "Axiom Space",
     "relation": "operator",
     "source": "axiom-spacebilt-axdcu",
     "url": "https://www.axiomspace.com/release/axiom-space-spacebilt-announce-orbital-data-center-node",
     "source_title": "Axiom Space, Spacebilt Announce Orbital Data Center Node Aboard International Space Station",
     "publisher": "Axiom Space",
     "source_kind": "press-release",
     "published": "2025-09-16",
     "published_basis": "dateline of the release",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in part: about six excerpts of 10,500 characters"
    },
    {
     "id": "c2",
     "kind": "quote",
     "text": "The prototype will test applications in cloud computing, artificial intelligence and machine learning(AI/ML), data fusion and space cybersecurity.",
     "speaker": "Axiom Space and Red Hat",
     "relation": "operator",
     "source": "axiom-redhat-axdcu",
     "url": "https://www.axiomspace.com/release/red-hat-teams-up-with-axiom-space-to-launch-optimize-axiom-spaces-data-center-unit-1-on-orbit",
     "source_title": "Red Hat Teams Up with Axiom Space to Launch, Optimize the Space Company's Data Center Unit-1 On Orbit",
     "publisher": "Axiom Space and Red Hat",
     "source_kind": "press-release",
     "published": "2025-03-06",
     "published_basis": "dateline of the release",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in part: about five excerpts of 4,600 characters"
    }
   ],
   "conflicts": [],
   "notes": "The second quotation is a plan, written before launch.",
   "changed": "Raised from L0 to L1: the operator's own release was read, where before the entry rested on another tracker.",
   "headline": "L1"
  },
  {
   "id": "kepler-relay-tranche-1",
   "entry": 16,
   "name": "Kepler optical relay satellites, first tranche",
   "group": "launched",
   "class": "compute",
   "location": "Earth orbit",
   "operator": "Kepler Communications",
   "developer": "Kepler Communications",
   "host": "Ten relay satellites",
   "catalogue": null,
   "dates": {
    "launched": "2026-01-11",
    "ran": null,
    "reported": "2026-03-16",
    "ended": null
   },
   "hardware": "40 NVIDIA Jetson Orin modules across ten satellites",
   "software": "none named",
   "did": "Launched with computing modules aboard, which the operator reported commissioned two months later and offers to customers. No AI workload is named as run in what was read.",
   "marks": [
    {
     "stage": "L",
     "evidence": 1,
     "basis": [
      "c1",
      "c2"
     ]
    }
   ],
   "claims": [
    {
     "id": "c1",
     "kind": "quote",
     "text": "multi-GPU on-orbit compute modules",
     "speaker": "Kepler Communications",
     "relation": "operator",
     "source": "kepler-first-tranche",
     "url": "https://kepler.space/kepler-successfully-launches-first-tranche-of-optical-relay-satellites/",
     "source_title": "Kepler successfully launches first tranche of optical relay satellites",
     "publisher": "Kepler Communications",
     "source_kind": "press-release",
     "published": "2026-01-11",
     "published_basis": "page metadata",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in full"
    },
    {
     "id": "c2",
     "kind": "quote",
     "text": "will now begin commissioning",
     "speaker": "Kepler Communications",
     "relation": "operator",
     "source": "kepler-first-tranche",
     "url": "https://kepler.space/kepler-successfully-launches-first-tranche-of-optical-relay-satellites/",
     "source_title": "Kepler successfully launches first tranche of optical relay satellites",
     "publisher": "Kepler Communications",
     "source_kind": "press-release",
     "published": "2026-01-11",
     "published_basis": "page metadata",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in full"
    },
    {
     "id": "c3",
     "kind": "quote",
     "text": "the commissioning of distributed on-orbit computing across its Tranche 1 optical data relay constellation",
     "speaker": "Kepler Communications",
     "relation": "operator",
     "source": "kepler-nvidia-cloud",
     "url": "https://kepler.space/kepler-deploys-first-space-based-scalable-cloud-infrastructure-powered-by-nvidia/",
     "source_title": "Kepler Deploys First Space-Based, Scalable Cloud Infrastructure Powered by NVIDIA",
     "publisher": "Kepler Communications",
     "source_kind": "press-release",
     "published": "2026-03-16",
     "published_basis": "dateline of the release",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in part: about six sentences of 6,000 characters"
    },
    {
     "id": "c4",
     "kind": "quote",
     "text": "powered by 40 NVIDIA Jetson Orin modules, deployed as distributed edge compute GPUs across 10 satellites",
     "speaker": "Kepler Communications",
     "relation": "operator",
     "source": "kepler-nvidia-cloud",
     "url": "https://kepler.space/kepler-deploys-first-space-based-scalable-cloud-infrastructure-powered-by-nvidia/",
     "source_title": "Kepler Deploys First Space-Based, Scalable Cloud Infrastructure Powered by NVIDIA",
     "publisher": "Kepler Communications",
     "source_kind": "press-release",
     "published": "2026-03-16",
     "published_basis": "dateline of the release",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in part: about six sentences of 6,000 characters"
    },
    {
     "id": "c5",
     "kind": "quote",
     "text": "The company now has 18 customers",
     "speaker": "TechCrunch",
     "relation": "news",
     "source": "techcrunch-kepler-cluster",
     "url": "https://techcrunch.com/2026/04/13/the-largest-orbital-compute-cluster-is-open-for-business/",
     "source_title": "The largest orbital compute cluster is open for business",
     "publisher": "TechCrunch",
     "source_kind": "news",
     "published": "2026-04-13",
     "published_basis": "page metadata",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in part: about five sentences of 4,400 characters"
    }
   ],
   "conflicts": [
    {
     "summary": "Another tracker lists this operator's compute cluster as commercially open since April 2026. The operator's release of March 2026 and a news report of April 2026 were then read: the computers are commissioned and have customers.",
     "status": "resolved: the computers are in service; no AI workload is named"
    }
   ],
   "notes": "Listed as launched, not as run, by a narrow reading: a computer in service with customers is not a named AI workload reported to have run. A customer's own account of what it ran there would change this.",
   "changed": "On 4 October 2026 two later sources were read. The mark is unchanged; the record no longer says the computers are untested.",
   "headline": "L1"
  },
  {
   "id": "suncatcher-prototype",
   "entry": 15,
   "name": "Project Suncatcher prototype",
   "group": "launched",
   "class": "compute",
   "location": "Earth orbit",
   "operator": "Google, with Planet",
   "developer": "Google",
   "host": "A prototype satellite built with Planet",
   "catalogue": null,
   "dates": {
    "launched": "2026-10-01",
    "ran": null,
    "reported": "2026-10-01",
    "ended": null
   },
   "hardware": "Google TPUs; the number aboard is not given in Google's posts",
   "software": "none reported",
   "did": "Reached orbit and made contact. Google says it will measure how its chips stand up to launch, radiation and heat; it reports no AI workload.",
   "marks": [
    {
     "stage": "L",
     "evidence": 1,
     "basis": [
      "c1"
     ]
    }
   ],
   "claims": [
    {
     "id": "c1",
     "kind": "quote",
     "text": "launched into orbit aboard the Transporter-18 rideshare mission with SpaceX. Our team has confirmed contact with the satellite and it is operating as expected",
     "speaker": "Google",
     "relation": "operator",
     "source": "google-suncatcher-in-orbit",
     "url": "https://blog.google/innovation-and-ai/models-and-research/google-research/project-suncatcher-prototype/",
     "source_title": "Our Project Suncatcher prototype satellite is in orbit.",
     "publisher": "Google",
     "source_kind": "blog",
     "published": "2026-10-01",
     "published_basis": "page metadata",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in full"
    },
    {
     "id": "c2",
     "kind": "quote",
     "text": "Over the coming weeks, we'll gather in-orbit data on how our TPUs handle the physical stress of spaceflight and the radiation and thermal extremes of space.",
     "speaker": "Google",
     "relation": "operator",
     "source": "google-suncatcher-in-orbit",
     "url": "https://blog.google/innovation-and-ai/models-and-research/google-research/project-suncatcher-prototype/",
     "source_title": "Our Project Suncatcher prototype satellite is in orbit.",
     "publisher": "Google",
     "source_kind": "blog",
     "published": "2026-10-01",
     "published_basis": "page metadata",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in full"
    },
    {
     "id": "c3",
     "kind": "quote",
     "text": "a pathfinder spacecraft for Project Suncatcher",
     "speaker": "Space.com",
     "relation": "news",
     "source": "spacecom-transporter-18",
     "url": "https://www.space.com/space-exploration/satellites/spacex-google-project-suncatcher-ai-satellite-transporter-18-mission",
     "source_title": "SpaceX launches Google AI satellite, 129 other payloads on 2nd leg of spaceflight tripleheader",
     "publisher": "Space.com",
     "source_kind": "news",
     "published": "2026-10-01",
     "published_basis": "page metadata",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in full"
    }
   ],
   "conflicts": [
    {
     "summary": "Three other trackers and round-ups, read on 3 October 2026, still gave this prototype as due in early 2027.",
     "status": "they are out of date"
    }
   ],
   "notes": "News reports give four TPUs; that figure was seen only in search snippets. Google's paper in Joule could not be opened. No catalogue object was found under the name tried.",
   "changed": null,
   "headline": "L1"
  },
  {
   "id": "moi-1a",
   "entry": 18,
   "name": "MOI-1A",
   "group": "launched",
   "class": "compute",
   "location": "Earth orbit",
   "operator": "TakeMe2Space",
   "developer": "TakeMe2Space",
   "host": "MOI-1A",
   "catalogue": null,
   "dates": {
    "launched": "2026-10",
    "ran": null,
    "reported": "2026-10-02",
    "ended": null
   },
   "hardware": "117 TOPS of compute, by the operator's description",
   "software": "none reported",
   "did": "Was launched at the start of October 2026, by a news report. Its operator calls it an orbital AI lab; nothing has been reported run on it.",
   "marks": [
    {
     "stage": "L",
     "evidence": 0,
     "basis": [
      "c2"
     ]
    }
   ],
   "claims": [
    {
     "id": "c1",
     "kind": "quote",
     "text": "An orbital AI lab with 117 TOPS of compute, 9-band multispectral imaging, and the power budget to run continuously—launching October 2026.",
     "speaker": "TakeMe2Space",
     "relation": "operator",
     "source": "tm2-moi",
     "url": "https://www.tm2.space/moi-satellite",
     "source_title": "LEO Satellite Infrastructure for Orbital Computing",
     "publisher": "TakeMe2Space",
     "source_kind": "product-page",
     "published": "undated",
     "published_basis": "no date of writing on the page",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in part: three excerpts of 5,000 characters"
    },
    {
     "id": "c2",
     "kind": "quote",
     "text": "TakeMe2Space launched its MOI-1a satellite on Friday morning aboard SpaceX's Transporter-18 mission",
     "speaker": "The Print",
     "relation": "news",
     "source": "theprint-moi-1a",
     "url": "https://theprint.in/feature/takeme2space-launches-ai-powered-satellite-on-spacex-eyes-orbital-data-centre/3059872",
     "source_title": "TakeMe2Space launches AI-powered satellite on SpaceX, eyes orbital data centre",
     "publisher": "The Print",
     "source_kind": "news",
     "published": "2026-10-02",
     "published_basis": "page metadata",
     "accessed": "2026-10-04",
     "copy_route": "wayback",
     "read": "in part: four sentences of 4,100 characters, from an Internet Archive capture"
    },
    {
     "id": "c3",
     "kind": "quote",
     "text": "The MOI-1a mission is a step towards building an orbital data centre in space.",
     "speaker": "The Print",
     "relation": "news",
     "source": "theprint-moi-1a",
     "url": "https://theprint.in/feature/takeme2space-launches-ai-powered-satellite-on-spacex-eyes-orbital-data-centre/3059872",
     "source_title": "TakeMe2Space launches AI-powered satellite on SpaceX, eyes orbital data centre",
     "publisher": "The Print",
     "source_kind": "news",
     "published": "2026-10-02",
     "published_basis": "page metadata",
     "accessed": "2026-10-04",
     "copy_route": "wayback",
     "read": "in part: four sentences of 4,100 characters, from an Internet Archive capture"
    }
   ],
   "conflicts": [
    {
     "summary": "The operator's own page, as fetched on 4 October 2026, still gave the launch as upcoming. A news report of 2 October says it launched. The catalogue has no object of that name yet.",
     "status": "the operator's page is taken to be out of date; the launch rests on news alone"
    }
   ],
   "notes": "The launch mark is 0: a news report, with nothing yet from the operator or the catalogue. The report gives the day as Friday morning, which is 2 October in India; the flight left on 1 October by universal time, so the date is recorded to the month.",
   "changed": "Published at L0 on 4 October 2026. It had been held back earlier that day, when the one news report could not be opened.",
   "headline": "L0"
  }
 ],
 "held_back": [
  {
   "id": "axiom-odc-nodes",
   "entry": 17,
   "name": "Axiom's two free-flying orbital data centre nodes",
   "group": "held-back",
   "class": "compute",
   "location": "Earth orbit",
   "operator": "Axiom Space",
   "developer": "Axiom Space",
   "host": "Kepler Communications relay satellites",
   "catalogue": null,
   "dates": {
    "launched": null,
    "ran": null,
    "reported": "2025-04-07",
    "ended": null
   },
   "hardware": "not stated",
   "software": "none reported",
   "did": "Announced for launch. No source read says the nodes are in orbit.",
   "marks": [
    {
     "stage": "A",
     "evidence": 1,
     "basis": [
      "c1"
     ]
    }
   ],
   "claims": [
    {
     "id": "c1",
     "kind": "quote",
     "text": "announced today the upcoming launch of its first two Orbital Data Center (ODC) nodes to low-Earth orbit (LEO), by the end of this year",
     "speaker": "Axiom Space",
     "relation": "operator",
     "source": "axiom-odc-nodes-announce",
     "url": "https://www.axiomspace.com/release/axiom-space-to-launch-orbital-data-center-nodes-to-support-national-security-commercial-international-customers",
     "source_title": "Axiom Space to Launch Orbital Data Center Nodes to Support National Security, Commercial, International Customers",
     "publisher": "Axiom Space",
     "source_kind": "press-release",
     "published": "2025-04-07",
     "published_basis": "dateline of the release",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in part: about five excerpts of 6,700 characters"
    }
   ],
   "conflicts": [
    {
     "summary": "A search-engine snippet of Axiom's product page says the nodes launched on 11 January 2026, and three trackers say the same. The text of that page as fetched holds no such sentence.",
     "status": "not supported by anything read"
    }
   ],
   "notes": "Held back: on what was read this is an announcement, and announcements are not listed in this version.",
   "changed": "Held back. In the first table it was listed as launched, on a snippet and on other trackers.",
   "headline": "A1"
  },
  {
   "id": "space-llama",
   "entry": null,
   "name": "Space Llama, for the International Space Station",
   "group": "held-back",
   "class": "language",
   "location": "space station",
   "operator": "Booz Allen Hamilton",
   "developer": "Booz Allen Hamilton, with Meta",
   "host": "International Space Station National Laboratory",
   "catalogue": null,
   "dates": {
    "launched": null,
    "ran": null,
    "reported": "2025-04-25",
    "ended": null
   },
   "hardware": "NVIDIA graphics processors, by the release",
   "software": "A fine-tuned version of Meta's Llama 3.2",
   "did": "Announced as a language model for the station's crew. The sentences read do not say in plain words that it has run there.",
   "marks": [
    {
     "stage": "A",
     "evidence": 1,
     "basis": [
      "c1"
     ]
    }
   ],
   "claims": [
    {
     "id": "c1",
     "kind": "quote",
     "text": "the development and successful demonstration of a novel AI-powered tech stack, accelerated by NVIDIA CUDA GPUs",
     "speaker": "Booz Allen Hamilton",
     "relation": "developer",
     "source": "boozallen-space-llama",
     "url": "https://newsroom.boozallen.com/news-releases/news-release-details/booz-allen-and-meta-launch-space-llama",
     "source_title": "Booz Allen and Meta Launch Space Llama",
     "publisher": "Booz Allen Hamilton",
     "source_kind": "press-release",
     "published": "2025-04-25",
     "published_basis": "date given in search results",
     "accessed": "2026-10-04",
     "copy_route": "wayback",
     "read": "in part: about eight sentences of 8,400 characters, from an Internet Archive capture"
    },
    {
     "id": "c2",
     "kind": "quote",
     "text": "Booz Allen has collaborated with Meta to deploy a fine-tuned version of Llama 3.2 aboard the International Space Station (ISS) National Laboratory",
     "speaker": "Meta",
     "relation": "model-maker",
     "source": "meta-space-llama",
     "url": "https://about.fb.com/news/2025/04/space-llama-metas-open-source-ai-model-heading-into-orbit/",
     "source_title": "Space Llama: Meta's Open Source AI Model Is Heading Into Orbit",
     "publisher": "Meta",
     "source_kind": "blog",
     "published": "2025-04-25",
     "published_basis": "date given in search results",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in part: three sentences of 4,700 characters"
    },
    {
     "id": "c3",
     "kind": "quote",
     "text": "to not only replace paper documents onboard the ISS National Lab, but also decrease reliance on instructions received from the ground",
     "speaker": "Booz Allen Hamilton",
     "relation": "developer",
     "source": "boozallen-space-llama",
     "url": "https://newsroom.boozallen.com/news-releases/news-release-details/booz-allen-and-meta-launch-space-llama",
     "source_title": "Booz Allen and Meta Launch Space Llama",
     "publisher": "Booz Allen Hamilton",
     "source_kind": "press-release",
     "published": "2025-04-25",
     "published_basis": "date given in search results",
     "accessed": "2026-10-04",
     "copy_route": "wayback",
     "read": "in part: about eight sentences of 8,400 characters, from an Internet Archive capture"
    }
   ],
   "conflicts": [
    {
     "summary": "News reports and trackers describe this as a model on the space station. The two announcements, as far as they were read, speak of a demonstration and of deploying; neither says where the demonstration ran.",
     "status": "not settled; both announcements were read only in part"
    }
   ],
   "notes": "Held back until the announcements are read in full. It may well have run in orbit.",
   "changed": "Added on 4 October 2026 and held back.",
   "headline": "A1"
  }
 ],
 "adjacent": [
  {
   "id": "perseverance-ai-drive",
   "entry": null,
   "name": "Perseverance's drives planned by AI",
   "group": "adjacent",
   "class": "planning",
   "location": "planned on Earth, driven on Mars",
   "operator": "NASA Jet Propulsion Laboratory",
   "developer": "NASA JPL, with Anthropic",
   "host": "Perseverance",
   "catalogue": null,
   "dates": {
    "launched": null,
    "ran": "8 and 10 December 2025",
    "reported": "2026-01-30",
    "ended": null
   },
   "hardware": "on Earth",
   "software": "Vision-language models (Anthropic's Claude)",
   "did": "Planned the rover's route on Earth; the commands were checked and sent to Mars, and the rover drove them.",
   "marks": [],
   "claims": [
    {
     "id": "c1",
     "kind": "quote",
     "text": "has completed the first drives on another world that were planned by artificial intelligence",
     "speaker": "NASA JPL",
     "relation": "operator",
     "source": "jpl-perseverance-ai-drive",
     "url": "https://www.jpl.nasa.gov/news/nasas-perseverance-rover-completes-first-ai-planned-drive-on-mars/",
     "source_title": "NASA's Perseverance Rover Completes First AI-Planned Drive on Mars",
     "publisher": "NASA Jet Propulsion Laboratory",
     "source_kind": "press-release",
     "published": "2026-01-30",
     "published_basis": "page metadata",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in full"
    },
    {
     "id": "c2",
     "kind": "quote",
     "text": "verifying over 500,000 telemetry variables before sending commands to Mars",
     "speaker": "NASA JPL",
     "relation": "operator",
     "source": "jpl-perseverance-ai-drive",
     "url": "https://www.jpl.nasa.gov/news/nasas-perseverance-rover-completes-first-ai-planned-drive-on-mars/",
     "source_title": "NASA's Perseverance Rover Completes First AI-Planned Drive on Mars",
     "publisher": "NASA Jet Propulsion Laboratory",
     "source_kind": "press-release",
     "published": "2026-01-30",
     "published_basis": "page metadata",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in full"
    },
    {
     "id": "c3",
     "kind": "quote",
     "text": "in collaboration with Anthropic, using the company's Claude AI models",
     "speaker": "NASA JPL",
     "relation": "operator",
     "source": "jpl-perseverance-ai-drive",
     "url": "https://www.jpl.nasa.gov/news/nasas-perseverance-rover-completes-first-ai-planned-drive-on-mars/",
     "source_title": "NASA's Perseverance Rover Completes First AI-Planned Drive on Mars",
     "publisher": "NASA Jet Propulsion Laboratory",
     "source_kind": "press-release",
     "published": "2026-01-30",
     "published_basis": "page metadata",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in full"
    }
   ],
   "conflicts": [],
   "notes": "Outside the register and shown beside it, because it marks the boundary: AI used for a space mission, not AI that ran in space. It carries no mark.",
   "changed": null,
   "headline": ""
  },
  {
   "id": "cimon",
   "entry": null,
   "name": "CIMON, the astronaut assistant",
   "group": "adjacent",
   "class": "language",
   "location": "spoken to on the space station, answered from Earth",
   "operator": "German Aerospace Center (DLR)",
   "developer": "Airbus, with IBM",
   "host": "International Space Station",
   "catalogue": null,
   "dates": {
    "launched": null,
    "ran": "15 November 2018",
    "reported": "2018-11-16",
    "ended": null
   },
   "hardware": "a free-flying device on the station; the language AI ran in IBM's cloud on Earth",
   "software": "IBM Watson",
   "did": "Talked with an astronaut on the station, with the understanding of speech done on Earth over a satellite link.",
   "marks": [],
   "claims": [
    {
     "id": "c1",
     "kind": "quote",
     "text": "CIMON uses the Wi-Fi on the ISS to transmit data, making use of satellite connections to ground stations to establish an internet connection with the IBM Cloud.",
     "speaker": "Airbus",
     "relation": "developer",
     "source": "airbus-cimon",
     "url": "https://www.airbus.com/en/newsroom/press-releases/2018-11-crew-assistant-cimon-successfully-completes-first-tasks-in-space",
     "source_title": "Crew assistant CIMON successfully completes first tasks in space",
     "publisher": "Airbus",
     "source_kind": "press-release",
     "published": "2018-11-16",
     "published_basis": "date given in search results; the release itself describes 15 November 2018",
     "accessed": "2026-10-04",
     "copy_route": "wayback",
     "read": "in part: three sentences of 7,700 characters"
    },
    {
     "id": "c2",
     "kind": "quote",
     "text": "The heart of the AI for understanding speech is the IBM Watson system.",
     "speaker": "German Aerospace Center (DLR)",
     "relation": "operator",
     "source": "dlr-cimon",
     "url": "https://www.dlr.de/en/research-and-transfer/projects-and-missions/horizons/cimon",
     "source_title": "CIMON: astronaut assistance system",
     "publisher": "German Aerospace Center (DLR)",
     "source_kind": "project-page",
     "published": "2018-06-08",
     "published_basis": "page metadata",
     "accessed": "2026-10-04",
     "copy_route": "direct",
     "read": "in part: one sentence of 2,900 characters"
    }
   ],
   "conflicts": [],
   "notes": "Beside the register, like the drives planned for Perseverance: an AI used in space whose model ran on Earth. Whether any of its software ran a model on the device itself was not looked into.",
   "changed": "Added beside the register on 4 October 2026.",
   "headline": ""
  }
 ]
}
