Skip to content
MicroFire Atlas

NASA Space Apps 2026 · Flame in Freefall

Fire behaves differently
when nothing rises.

MicroFire Atlas turns NASA's microgravity combustion tests into traceable evidence for safer spacecraft, and shows exactly where that evidence stops.

Enter Challenge ModeExplore the Atlas

Every value links to its NASA page. Scenery is illustration; the evidence is real.

Choose your mission

One question. Nine steps. Every claim traceable.

What does NASA actually know about PMMA fire in a future lunar habitat? Challenge Mode answers end to end on one page, then hands you a Mission Evidence Brief.

Enter Challenge ModeJudge Mode: 3 demos and a guided tour
  1. 01Find
  2. 02Compare
  3. 03Summarize
  4. 04Rank
  5. 05Interpret
  6. 06AI
  7. 07Uncertainty
  8. 08Next experiment
  9. 09Traceability

Experience the evidence

Real NASA footage, computer-vision measurements, matched NASA tests and NASA's own findings, in one instrument.

Real NASA video
Saffire VI · NASA Image and Video Library · shown as NASA released it

NASA observation

This isn't animation. NASA burned this material in space.

Saffire VI, inside an uncrewed Cygnus cargo ship at the end of the NG-19 mission: a one-sided PMMA sample, filmed by NASA and shown at the speed NASA released it.

Experiment
Saffire VI
Material
PMMA, burning on one side
Not stated
Oxygen, pressure and airflow for this clip

NASA Image and Video Library ↗

Research Frontier

Where NASA's evidence stops

MicroFire Atlas refuses to fabricate conclusions when evidence stops. Every cell is counted from the real records: gravity against oxygen.

✓ 5 known△ 2 partial? 11 unknown

NASA test records by gravity and oxygen. Select a cell for its records or its gap.
GravityO₂ %< 1717–1919–21.521.5–2626–31≥ 31
Orbitmicrogravity
Moonlunar g · simulated only
MarsMartian g

Pins: A = Exploration atmosphere A · 34 % O₂ · 56.5 kPa; B = Alternate atmosphere · 28.5 % O₂ · 66.2 kPa; ISS = ISS cabin · 21 % O₂ · 101.3 kPa. Earth gravity is not a row: this atlas holds no 1-g tests, and NASA screens materials on Earth with a separate standard.

Moon · ≥ 31 % O₂ Unknown: data gap

A Exploration atmosphere A · 34 % O₂ · 56.5 kPa

No NASA test in this atlas burned here. MicroFire shows the closest evidence, names the experiment that would fill this cell, and gives no outcome.

What experiment would close this gap?

Evidence status

  • No NASA record covers the condition. MicroFire abstains: it shows the closest evidence, names the experiment that would close the gap, and the AI Model Lab blocks any prediction.

Limitations we state up front
  • A curated atlas of 78 records, not the complete NASA combustion archive.
  • Similarity, ranking and coverage scores are MicroFire heuristics, never NASA ratings.
  • No safe or unsafe ratings, risk scores or fire probabilities anywhere.
  • Flame Vision reports pixels; NASA publishes no calibration for the footage.
All limitations

How MicroFire thinks

  1. ObserveNASA test tables, footage and reports
  2. ExtractTyped records in NASA's own units and words
  3. ValidateEvery quote and number checked against the PDF
  4. CompareEvidence Ladder and ranking with stability checks
  5. ExplainAI synthesis, then claim-by-claim verification
  6. AbstainNo number where NASA has not tested

How MicroFire thinks

From a NASA test table to a bounded answer. AI is used where it helps and stopped where the evidence stops. 78 test records, 62 quoted findings, 19 NASA documents.

  1. 01NASA evidence78 test records, 62 quoted findings, 19 NASA documents
  2. 02Structured retrievalSearches typed records and checked quotes, never the open web
  3. 03Evidence rankingDirect, analogous or mechanistic, with a stability check on every rank
  4. 04AI synthesisA language model explains only the evidence it was handed
  5. 05Claim verificationCitations, numbers, units and causal wording checked after generation
  6. 06Mission interpretationYour cabin conditions against the closest NASA tests
  7. 07Uncertainty and abstentionOutside the evidence, the system says so and gives no number
  8. 08Research gapThe untested condition, written as the next research question

What the AI does, and does not do

  • Evidence retrieval

    Find relevant NASA tests across BASS-II, Saffire and LUCI in one search.

    Ask a question
  • AI synthesis

    Explain retrieved evidence in plain language. Live synthesis is paused on the public site; a saved answer is re-verified on every build.

    See a verified answer
  • Claim verification

    Every generated claim must cite retrieved evidence and keep its numbers, units and gravity context.

    Try to fool the checker
  • Evidence-bounded ML

    A validated model on 41 BASS-II tests that refuses to predict outside the conditions NASA tested.

    Open the Model Lab
Read the full methodology

Investigate deeper

Decades of NASA fire research sit in separate reports. Here a researcher can:

Enter Challenge ModeOpen the Flame Lab