# Four-minute judging pitch — model 1.1.0

Use the seven-slide PDF/PPTX and preload `/demo`. A spoken target of about 500 words leaves a little room for transitions. The demo runs the actual model; it does not load prerecorded outcomes. Keep `offline.html` and the PDF locally.

## 0:00–0:25 · Slide 1 · A consequential choice

“Your lunar outpost has two astronauts, a greenhouse and a solar array. Then a circuit fails. Would you pause science, stop caring for future food, or spend your remaining battery?

I’m Edmund Gunn Jr., and this is gunnchOS’s 7GC Mission Control, built for the Junior Astronaut Mission Trainer challenge.”

## 0:25–1:00 · Slides 2–3 · Play the emergency

“The learner becomes flight director for one complete 24-hour exercise. Oxygen generation spends water and electricity. Recovery saves water but draws power. The greenhouse competes for both.

At hour six, the array fault reduces power. The repair robot needs energy for two hours before generation returns. The player sees the next-hour budget before committing. During the radiation drill, shelter protects the crew and pauses experiments.”

Switch to `/demo`; click **Recovery plan** while narrating its eight-second run, then **Ignore fault** for the three-second failure. If time is short, use only the ignored-fault run and the actual results in slide 5.

“Leave everything unchanged and power fails at hour nine. Decisions change the resource flows and outcome.”

## 1:00–1:45 · Slide 5, or the live planner · Compare defensible plans

“Our signature feature compares strategies through the same simulation engine. It evaluates 567 phase policies against explicit reserve requirements. It exposes a frontier of competing objectives, rather than declaring one universally best answer.

A science-focused plan earns twelve science hours. A balanced plan trades some of that work for twelve hours of greenhouse care. A greenhouse-focused plan supports all twenty-four care hours but completes only four science hours.

These are computed outcomes. Crop care is an activity proxy, not a claim that we grow a new harvest in one day.”

On the demo page click **Compare strategies**. Show the default balanced result briefly, then the **Margins under all tested changes** preference. Do not read every row.

## 1:45–2:25 · Slide 4 · Sources and validity

“NASA’s published planning rates set crew oxygen, water and dry-food use. Electrolysis adds the physical connection: approximately 1.125 kilograms of water for each kilogram of oxygen generated.

We keep units consistent and record actual supply, demand, losses and unmet demand. Hardware capacities, greenhouse loads, exposure points and radio parameters are clearly labeled teaching assumptions. We model neither medical radiation dose nor a flight-qualified habitat.

One deterministic transition drives play, forecasts, planning and exact replay. Forty-one tests verify the model and planning behavior.”

## 2:25–3:00 · Slide 5 · Resilience has a cost

“We rerun fixed decisions with less solar power, greater crew use, a longer repair and their combination. Eighteen policies preserve the required margins across all five tested conditions.

Our robust choice gives up some activities to retain at least 1.035 kilowatt-hours in the combined case. The choices do not retune themselves after seeing the result.

A separate simulated radio outage buffers science data locally and drains the queue after recovery. Overflow appears as lost data. This is an inspectable queue model, not an operational lunar network.”

## 3:00–3:35 · Slide 6 · Accessible learning

“Junior mode offers friendly language and a dictionary. Explorer compares plans. Engineer exposes equations and traces. All three are views of the same mission.

Time waits for the learner. The application needs no account or API key, and its standalone HTML includes the planner and explanations offline. Browser checks cover mobile width, keyboard controls, exports and offline play.

Our educational goal is for learners to explain what their plan sacrificed and why. Educator feedback and a classroom pilot come next; we have not yet measured learning outcomes.”

## 3:35–4:00 · Slide 7 · Contribution and invitation

“I directed this project’s goals and engineering requirements, with OpenAI Codex assistance for implementation, writing and original visual design. The NASA references, source code, test evidence and limitations are available for inspection.

The existing gunnchOS production ecosystem remains separate. We want a learner to finish this mission able to explain why keeping one system alive sometimes means giving another activity less.”

## Manual emergency route

1. Reference conditions. Advance to hour 6.
2. Switch science off. Queue robotic repair. Forecast: solar 0.65 kW, load 2.55 kW, battery change −1.90 kWh/h. Two repair hours require 3.80 kWh from the battery under the fixed-load assumption.
3. Advance to repair completion at hour 8, then hour 10. Shelter the crew.
4. Advance through the hour-12 radio event to hour 14. Leave shelter and restore science if desired.
5. For the reference recovery, run science through hour 22, then pause it for the final two hours. It completes 14 science hours but touches zero battery at a boundary, missing the planner’s extra margin.
6. Prefer the planner’s **Start a fresh mission with this plan** for a reserve-feasible demonstration. Balanced gives 9 science and 12 crop-care hours, minimum battery 0.75 kWh.

## Questions for science and engineering judges

**Which values are NASA-derived?** Surface oxygen 0.84 kg/person/day and water components totaling 3.20 kg/person/day come from NTRS 20230012635 Table 8. Dry food is the non-EVA dry component, 0.98 kg/person/day, from Table 2; it is not total packaged food mass. Electrolysis uses rounded molecular stoichiometry consistent with Table 11. NASA ECLSS supports the system-coupling explanation. Hardware, greenhouse, radio, margins and risk values are authored.

**Why can a completed mission still fail a planning check?** Mission completion requires no unserved essential power and no exhausted usable stocks. Planning additionally requires reserves throughout the trajectory. A zero battery can coincide with adequate current solar generation, so it is distinct from a blackout; it still leaves no stored-energy margin.

**Is the planner globally optimal?** No. It is exhaustive within 567 phase specifications. It fixes life support/radio on and shelter during the drill, and varies repair timing, science windows/targets and greenhouse targets. Arbitrary hourly scheduling and radio duty cycling lie outside the grid. Preferences and tie breaks are documented.

**What does robust mean here?** Passing five explicit deterministic cases with the same decisions. It does not establish a failure probability, universal robustness or NASA certification.

**Do plants feed the astronauts today?** No. Stored dry food supplies them. Greenhouse care consumes water/power and earns an activity proxy; crop biology, nutrition and harvest are future extensions.

**What is original about the communications element?** The contribution is an inspectable, bounded queue coupled to the mission power budget, showing local buffering, recovery and overflow. It is simulated and does not claim a measured RF link or lunar 6G service.

**What is tested?** Forty-one model/planning tests, exact exported replay, plus real-browser manual decisions, learning-view invariance, computed strategy application, mobile layout, keyboard map, downloads and offline paths. These are software-behavior checks, not classroom or accessibility certification.

**What is omitted?** Explicit CO₂/cabin pressure, thermal dynamics, real shielding, efficiency/rate limits, crop growth, full habitat mass closure and physical RF propagation. Hourly aggregation does not resolve sub-hour events or medical outcomes.

**How was AI used?** OpenAI Codex assisted code, illustrative parameters, documentation, original SVG design, tests and presentation. Public labels and metadata identify that use. The runtime has no AI inference. The owner supplied the concept, constraints and project direction.

**What would you research next?** Begin with learner/educator and assistive-technology testing, then validated component loads and crop/thermal/CO₂ models, richer constrained scheduling, and a documented communications link/service model. Do not claim those extensions already exist.

## Operator constraints

Chicago local cutoff from owner: November 15 at 1:30 PM America/Chicago. Global cutoff: 11:59 PM local time. Owner reports no separate local link. Baseline was submitted at 4:09 AM CT; record the final resubmission confirmation separately. Leave travel/setup time. Official venue details list The Ferris, 1 West Monroe Street.