OpenAI Build Week 2026

O2 by Agent9.

A public air-quality dashboard that turns wildfire smoke, AQI readings, and forecast data into plain-language guidance: what the air is doing now, when it may improve, and how to plan around it.

Open O2 →
OpenAI CodexReactOpen-Meteo CAMSNOAA / NWS Smoke DataCloudflare

Make smoke data usable

Wildfire smoke information is scattered across maps, forecasts, agency bulletins, pollutant readings, and local alerts. That leaves people with a practical question that most dashboards do not answer directly: when will the air get better enough to go outside?


O2 by Agent9 was built for the OpenAI Build Week 2026 Hackathon to make that answer easier to find. The submitted project is also available on Devpost: O2 by Agent9.

  • Local Forecast A location-aware AQI readout with a 48-hour forecast and the cleanest upcoming window.
  • Smoke Context NOAA/NWS smoke data and Canadian FireSmoke BlueSky overlays help distinguish ground-level air quality from visible smoke aloft.
  • Plain Guidance Every AQI category is translated into direct guidance for everyone and for sensitive groups.
  • Alert Flow Opt-in air alerts are structured around unhealthy air and safe-again conditions.
The Solution

Forecast First

O2 centers the planning question instead of only showing a current AQI number, naming the expected improvement window when forecast data supports it.

Source Provenance

The interface lists its data sources and marks whether each signal is measured, modeled, or forecast so users understand what they are looking at.

Accessible Safety UI

Air-quality categories are written out in words as well as color, with keyboard-operable controls and text summaries for critical guidance.

Results
48h
AQI Forecast
2
Smoke Layers
2026
Build Week

NEED A PUBLIC DATA PRODUCT?

We build practical AI-assisted tools that turn complex data into decisions people can act on.

Start the Conversation →