Recon: NASA Space Apps Challenge 2026

14–15 November 2026 · hybrid (in-person and virtual Local Events + the virtual Universal Event) · theme “The Next Frontier!” · run 2026-09-14 · corpus slug nasa-space-apps-2026, prior nasa-space-apps-2025.

Intake assumptions (defaults — replace when the team is known)

Rules that change the harness

• Teams must not begin working on the challenges before the hackathon (organizer FAQ). Challenge summaries appear 2026-09-17 and full statements 2026-10-28. Until 14 November this plan can only produce recon, patterns, angle shapes and general preparation — no challenge-specific solution, design or code.
• Teams of at most six, all registered for the same Local Event (or all in the Universal Event).
• Submit through the team page on spaceappschallenge.org by the end of the Local Event (23:59 local time on the last day in 2025).

Rubric and lens weights

Each criterion is scored 1–5 at every judging level. Equal weights are an assumption: no percentages are published. The organizer’s own advice (2024 guide) is that teams do well on the first four criteria and under-invest in presentation.

Judging funnel

Implication: two of the three phases most likely judge the submission without the team in the room. The project page, demonstration link and project link are the pitch for the global phases; the live pitch matters only locally.

Prizes and lanes (2025 award definitions)

Stacking: in 2025 each award went to a different team, so plan for one award per team (inference). Local Awards are separate and do not qualify a team for global judging.

Constraints and submission requirements (2025 rules unless noted)

• Project page in English; a demo video in another language needs English subtitles.
• Use and link at least one NASA data source or resource.
• Disclose every AI tool used; watermark AI-generated images and video; acknowledge AI-generated code, data, audio and text.
• No NASA branding (logo, flag, mission identifiers).
• “Create Your Own Challenge” projects are not eligible for global judging.
• 2026: no work on challenges before the event; team size ≤ 6; submission guide published 2026-11-13.

Timeline

Saturation and past editions

• No base rates. The corpus has the Global Winners only (roster_coverage: winners-only): 10 of 11,500+ submissions in 2025 (18 challenges) and 10 of 9,996 in 2024 (20 eligible challenges). Nothing here says which lanes are crowded.
• Commercializing LEO produced 3 of the 2025 winners and the agriculture challenge 2 of the 2024 winners — a challenge can yield several awards; that does not mean it was uncrowded.
• LATAM teams won 3 of 10 awards in both editions (2025: Mexico, Brazil, Peru; 2024: all Brazil).

Domain data availability

NASA resources the 2025 winners named on their team pages: POWER API, Earthdata (Giovanni, GES DISC, earthaccess), DONKI space-weather database, GeneLab (OSD-288), MAST archive (Kepler, TESS, JWST, Hubble), Landsat and MODIS vegetation indices (NDVI, EVI, NDMI), FIRMS fire data, LAADS DAAC, the U.S. Drought Monitor, Orbital Debris Program Office tools (Debris Assessment Software, ORDEM), and the NASA Image and Video Library. Account setup and learning these APIs is general preparation, not challenge work.

LATAM public-sector data (via Croma) is only relevant if a 2026 challenge touches public services; not applicable yet.

Sources

• https://www.spaceappschallenge.org/resources/participant-faqs/ — retrieved 2026-09-14
• https://www.spaceappschallenge.org/2026/ — retrieved 2026-09-14
• https://www.spaceappschallenge.org/2025/awards/ — retrieved 2026-09-14
• https://www.nasa.gov/learning-resources/stem-engagement-at-nasa/nasa-announces-2025-international-space-apps-challenge-global-winners/ — retrieved 2026-09-14
• https://www.youtube.com/watch?v=SAm_2GB7PGc (official 2025 judging video) — retrieved 2026-09-14
• 2025 team pages linked from each project record — retrieved 2026-09-14

Patterns: what won at the last edition, and how it differs from other events

Queries: crafterwiki event nasa-space-apps-2026, patterns --event-kind agency-open-data-challenge, patterns --event-kind student-major-league, patterns --event-kind open-source-community, three similar profiles (see 03-angles.md). Corpus state 2026-09-14.

Sample

n = 10 confirmed winners (the 2025 Global Winners), one edition, one event. Mechanism tags are mostly curator inference; organizer- and team-stated claims are labelled in each record. Treat every number below as a prior, not a law.

Agency open-data challenge (NASA Space Apps 2025, n = 10)

• Build: software 8, concept design 1, media/story 1, custom hardware 0.
• AI in the product: 5 of 10 used none. The rest: LLM API 2, trained or custom ML model 2, computer vision 1, generative media 1, unspecified 1. AI was rarely the reason a project won.
• Submission package: all 10 team pages carry a demonstration link and a project link; 9 of 10 link a deployed site or repository (the concept winner linked slides and a folder).
• Team size: one solo winner, four teams of 4, one of 5, four of 6.

How this differs from other kinds of event

Reading: the student-event playbook (hardware on the table, a pop-culture hook, a live “wait, it does that?”) does not transfer. At the agency challenge, winners answered the challenge closely, made expert data usable by a named person or place, and showed real depth — often without AI.

Similar situations

Every shape has a 2025 winner as its nearest neighbour: these shapes can win, and judges have seen them. The edge has to come from combination and specificity, not from the shape.

Crowded lanes

Unknown. With a winners-only roster the corpus cannot say how many of 11,500+ submissions built a weather app or a farming game. Curator inference: “personal weather/climate app” is the most obvious reading of Earth-observation challenges, so assume it is common and plan an edge.

Update: two editions (2026-09-14)

The 2024 Global Winners were indexed as a holdout after this file was written (docs/research/holdout-nasa-2024.md). Agency sample now n = 20.

Reading: accessible expert data and visible depth held; human stakes did not. The angle verdict in 03-angles.md is revised accordingly.

Pre-registered predictions for the 2026 Global Winners

Written 2026-09-14, before 2026 challenges exist. Grade in January 2027 by tagging the 2026 winners with the same taxonomy from NASA’s announcement and team pages. Baseline in brackets is the 2025 value.

P7–P8 were added on 2026-09-14 after the 2024 holdout and before any 2026 challenge exists; P1–P6 are unchanged. P7–P8 test the award-first step proposed in 03-angles.md.

Context prior (not a harness claim): LATAM teams won 3 of 10 awards in 2025.

Problems and angle shapes

Rule-limited phase. Teams may not work on 2026 challenges before 14 November, and the challenges are not public yet. This file holds problem families and structurally different angle shapes keyed to the award lanes — not solutions. On 2026-09-17 map each family to a challenge summary; drop families with no match.

Problem families (≥ 6)

Killed now: any family that needs custom hardware (no 2025 precedent and no hardware assumed), and any “platform for everything” without a named person.

Evidence pass

• search "weather" → hercode-space, pureflow, skysense (all award winners).
• search "satellite" → photonics-odyssey, zumorroda-x, quenaris (all award winners).
• No base rates exist for this event (winners-only roster), so crowding is unknown; assume families 1 and 4 are common readings of Earth-observation challenges.

Angle shapes

A · One-tap local decision

B · Stress test with live data

C · Play the stakeholder

Judge panel (Gate 1)

Personas

1. Local judge (mentor at a LATAM Local Event) — rewards a clear pitch, a working demo, local relevance.
2. Earth-science SME — rewards correct data use, stated uncertainty, a checkable method.
3. Heliophysics / space-operations SME — rewards physical plausibility.
4. Executive committee member — rewards stakes and memorability among 45 finalists.
5. Science communicator — rewards a page and video that explain it in a minute without the team.

Scores (1–5, equal weights)

Seen-before check

All three shapes have a 2025 award winner as nearest neighbour (skysense, quenaris; resonant-exoplanets, pureflow; zumorroda-x, hercode-space). That proves the shapes win, and it means the SMEs have seen them. Crowding among submissions is unknown (winners-only roster).

Biggest objections

Verdict

• MERGE A + B → “Local decision with a live validity check.” One place, one person, one decision from NASA data, plus a visible check of when that decision should not be trusted. Keeps A’s relevance and stakes; borrows B’s validity edge.
• KEEP C in reserve only if the chosen 2026 challenge asks for a story or a game and the team has an illustrator.
• KILL B standalone: its relevance depends on a narrow set of challenges.

Revision after the 2024 holdout (2026-09-14)

The verdict above was tested on the 10 NASA Space Apps 2024 Global Winners, indexed after the hypotheses were written (docs/research/holdout-nasa-2024.md).

• Held across both editions: expert data made usable by non-experts (12 of 20 winners), visible engineering depth (10 of 20), no custom hardware (0 of 20), AI optional (9 of 20 without AI).
• Did not hold: decision tools were 2 of 10 2024 winners and validated outputs 2 of 10; human-stakes fell from 5 of 10 to 1 of 10. “Local decision with a live validity check” is one option, not the direction.
• New step on 2026-09-17 and 2026-10-28: choose the target award before the shape, then score shapes against that award’s definition as well as the rubric.

Revised verdict

• Default for Earth-observation challenges aiming at Local Impact or Galactic Impact: A + B as above.
• Promote C from reserve to a first-class option when the team targets storytelling, art or inspiration — no longer conditional on the challenge asking for a game.
• B’s validity check is a differentiator for science and impact awards, not a requirement.
• The award table is post hoc with two winners per award; it is pre-registered as P7–P8 in 02-patterns.md and graded in January 2027.

Demo script: local decision with a live validity check

Template for the merged angle. Bracketed parts are filled on 14 November, once the challenge is chosen.

One sentence

“For [person] in [place], we turn NASA [dataset] into [one decision] — and show when not to trust it.”

Repeatability test: a judge should be able to say it to another judge without the words API, model or dashboard.

Local Event pitch (default 3:00; use the Local Event’s format)

Real vs mocked

Fallback

• Cached data snapshot for the demo region, switched on by the demo driver if the network or the feed fails.
• Recorded screen capture of the moment, labelled as a recording if used.
• A second laptop with the build open; a second person who can present.

Stage plan

• Speaker: story/pitch lead. Driver: app lead. Data lead answers validity questions.
• Nothing on the table: this event does not reward props (0 of 20 hardware winners in 2024–2025).

Global judging variant (the submission wins without you)

Mirror the section order every 2025 winner used on its team page:

1. Summary — first line is the one sentence; then person, place, decision, and why the data can be trusted.
2. Project Demonstration — link to a short video that opens on the decision (length and format per the 2026 Project Submission Guide, published 2026-11-13; do not assume).
3. Project — live site or repository that still works after the event.
4. Project Details — how it works, the validity check, tools, what is next.
5. Use of Artificial Intelligence — tool by tool, what each did and what the team built (the 2025 Best Use of Technology winner is a good model).
6. NASA Data — every dataset, named and linked.
7. Space Agency Partner & Other Data — everything else, named and linked.

Checklist: English throughout (subtitles for other languages) · at least one NASA resource linked · AI disclosure and watermarks · no NASA branding · all members registered at the same Local Event.

Rehearsal checklist

• Two full timed runs before submission; cut anything that runs long.
• One run with the network off to test the fallback.
• A teammate who did not build it reads the page cold and says the sentence back.

Architecture: Gate 2 is closed

Status: not approved, by design. Gate 2 needs the team’s approval of a concrete angle, and the event forbids working on challenges before 14 November; the challenge statements arrive on 28 October. This file therefore records only the constraints any demo-path architecture must meet. The actual design is made at the start of the hackathon, after the team approves the angle.

Demo path (shape only)

flowchart LR
  Q[Place + date + question] --> F[Fetch NASA dataset for the region]
  F --> C[(Cached snapshot for the demo region)]
  F --> D[Deterministic decision logic]
  C --> D
  D --> V[Validity check: agreement, staleness, coverage]
  V --> O[Decision + evidence + confidence]

Constraints decided now

1. Only the demo path gets built. One region, one person, one decision. Everything else is a slide.
2. No language model in the live decision path. If an LLM is used, it only phrases the explanation, a deterministic answer exists without it, and it is disclosed.
3. A cached snapshot for the demo region ships with the build; the live fetch is attempted first.
4. The team’s known stack, chosen in the first 30 minutes of the hackathon — no stack debates.
5. Evidence travels with the answer: dataset name, date and link are part of the output, not a footnote.
6. Validity check is a component, not a slide: it must run on the demo query.

Spikes for the first two hours of the hackathon

Build plan

Two parts: legitimate preparation until 13 November (no challenge work), and the plan for the hackathon itself.

Part 1 — Preparation (now → 13 November)

Part 2 — Hackathon (relative hours; fit to the Local Event schedule)

Roles from hour one

With 5–6 people add a science reviewer (checks every claim on the page against the data) and a video editor.

Bottleneck check

After the event

Run wiki-curate to record planned vs actual: which angle shape was used, what changed at Gate 2, and the outcome at each judging phase. Grade the predictions in 02-patterns.md when the 2026 Global Winners are announced.