OpenAI Build Week
8 winning projects reviewed
A concise independent look at what the winning projects actually built.
I also participated in this hackathon with Trajectory. This coverage was prepared independently after results were announced.
What we found
A few patterns from the winning projects we reviewed.
What AI did, and what stayed humanWorking artifacts made the strongest connections: Mechanica’s local build matched its deployment byte for byte, and Sentinel’s release digest matched its README.
AI had a bounded job in several winners. AirBridge checks assistant actions against a local allowlist; Echo Canvas builds its audio graph entirely in the browser.
Building with AI did not always mean putting a language model in the finished product. Pulse’s runtime uses speech-to-text, while veTriage’s shipped routing is deterministic.
Access and hardware set the remaining limits. Second Voice’s confirmation flow stayed unconfirmed without its implementation, and AirBridge’s timing log still needs an independent receiver test.
Want this before winners are announced?
Finalist verification can run privately before awards are announced, giving organizers time to resolve open questions.
About the hackathon
A global, online week of building with Codex. Participants had nine days to create a working project with Codex and GPT-5.6, an app, agent, website, game, or workflow, in one of four tracks, and to show where Codex accelerated the work. Submissions needed a demo video under three minutes, a repository with a README (public or privately shared with the judges), and the Codex session ID.
- Participants
- 46,703
- Submissions
- 8,015
- Tracks
- Apps for your life · Work and productivity · Developer tools · Education
- Prizes
- $100,000 in total: $15,000 for first and $10,000 for second place in each track. First-place winners also received Dev Day passes, promotion by OpenAI, a meeting with the Codex team, and a year of ChatGPT Pro.
- Requirements
- Built with Codex and GPT-5.6; demo video under three minutes; repository with README, public or privately shared; Codex session ID.
- Judged on
- Technological implementation, Design, Potential impact, Quality of the idea
- Judges
- Thibault Sottiaux, Kath Korevec, Tara Seshan, Leah Belsky, Peter Steinberger
From the official event page, checked .
The winners
-
Second Voice
Ravitez Dondeti
A builder-described communication aid for dysarthric speech: the builder describes candidate sentences and confirmation before speech output. We could not independently confirm the implementation or zero-training claim.
Apps for Your Life
1st place
-
AirBridge for Windows
Adam Tarantino
A Windows-to-AirPlay audio bridge with a file-free streaming path and a policy-gated assistant in the code. Speaker alignment is implemented, but we did not test it on hardware.
Apps for Your Life
2nd place
-
veTriage
Erin Downes VMD
A four-color phone-triage prototype for veterinary receptionists. The deployed staging build uses rules for triage decisions and labels itself as testing only, not approved for clinical use.
Work & Productivity
1st place
-
Pulse
Mohamed Mostafa
A cardiac-arrest research prototype with a deterministic engine for rhythm, shocks, and drug timings. Our runs used simulated transcription; live room-audio recognition was not exercised.
Work & Productivity
2nd place
-
Echo Canvas
Kevin Yang
A browser workbench for sketching rooms and sound sources, with HRTF spatial audio observed in the deployed app. We did not exercise its model-generated scene flow.
Developer Tools
1st place
-
Sentinel
Malik Bashaar Javaid
An MCP security scanner whose submission code combines seven static rules, a constrained model review, and four Docker probes, with findings mapped to the OWASP Agentic Top 10.
Developer Tools
2nd place
-
Mechanica
Weiying Zhu, Yukun Li, Shan Wei
Four ancient Chinese machines presented as interactive, source-cited 3D reconstructions, with a museum-grounded docent. The mechanisms are kinematic models; our checks found geometry and citation-rendering limitations.
Education
1st place
-
Dấu
Robert Huynh
A Vietnamese tone lab with pitch extraction and reference-contour comparison in the shipped browser code. Its server-side coaching was not exercised, and accuracy figures remain the builder’s measurements.
Education
2nd place