186 Countries, 47,000 Builders: Meet the 8 Winners of OpenAI Build Week!

Hey everyone, it's Shiichan! Today I've got news on a genuinely huge scale — a story about developers all over the world building things together.

OpenAI Developers developers.openai.com

What was announced?

On OpenAI's Developer blog, the winners of "OpenAI Build Week" were announced. Build Week was a challenge to build practical projects using Codex and GPT-5.6, and it drew around 47,000 builders from 186 countries and territories, with more than 8,000 projects submitted. It's reportedly one of the largest hackathons ever held.

Why it matters

That scale of participation, from that many countries at once, is real evidence of how broadly Codex and GPT-5.6 can support people's "I want to build this" instinct. What stands out is that it wasn't just professional engineers — people with zero coding experience took part too.

What changes

Looking at the winning projects gives you a really concrete sense of what's actually buildable with AI coding tools today. From accessibility, to decision support in medical settings, to developer security tools, to educational content, the genres are all over the map — but every one of them uses Codex or GPT-5.6 to solve a real problem. If you want to join a challenge like Build Week yourself, or you're just getting started with Codex, these make for solid inspiration.

Dive Deep

Winners were chosen across 4 categories, 8 projects in total. The prize pool totaled $100,000, and each category's 1st-place winner receives access to OpenAI's DevDay, a meeting with the Codex team, and a year of ChatGPT Pro.

Apps for Your Life

• 1st: "Second Voice" (by Ravitez Dondeti) — a speech-assistance tool for people with dysarthria or motor-control difficulties. It combines partial speech, saved phrases, and context to suggest sentence candidates, which the user selects and edits before it's spoken aloud
• 2nd: "AirBridge for Windows" (by Adam Tarantino) — streams audio from a Windows PC to HomePods, Apple TVs, and Macs. It supports multiple speakers, per-room latency calibration, and video lip-sync correction, and can be controlled by a GPT-5.6-powered voice assistant

Work & Productivity

• 1st: "veTriage" (by Erin Downes, DVM) — a triage app for veterinary clinic front desks that lets non-medical staff separate symptom urgency from hospital capacity without making a medical judgment call. It was built by a 61-year-old vet with zero prior coding experience, using Codex and multimodal models
• 2nd: "Pulse" (by Mohamed Mostafa Mohamed Labib Abu Taleb) — a medical-support prototype for cardiac-arrest response scenarios. GPT-5.6 interprets speech that code-switches between Egyptian Arabic and English, while deterministic code tracks and verifies the resuscitation workflow

Developer Tools

• 1st: "Echo Canvas" (by Kevin Yang) — a visualization tool for spatial audio design that computes real-time acoustics in the browser as you sketch spaces, place sound sources, and change materials. GPT-5.6 only operates within a constrained schema, while geometry calculations are handled by DSP and a physics engine
• 2nd: "Sentinel" (by Malik Bashaar Javaid) — a tool that inspects MCP servers for security vulnerabilities, combining deterministic static analysis, context-limited GPT-5.6 review, and Docker-isolated probes, then maps results to the OWASP Agentic Top 10

Education

• 1st: "Mechanica" (by Weiying Zhu, Yukun Li, and Shan Wei) — an interactive digital reconstruction of ancient Chinese machines, like astronomical clocks and programmable looms, that clearly separates historical record, measured data, and scholarly inference. Codex helped implement the 3D modeling and physics simulation, while GPT-5.6 acts as an "AI curator" presenting evidence
• 2nd: "Dấu" (by Robert Huynh) — a visual coach for learning Vietnamese tone, overlaying a user's pronunciation curve against a certified native-speaker baseline. Deterministic signal processing handles tone scoring, while GPT-5.6 handles the coaching

One thing all these projects have in common: instead of handing GPT-5.6 everything, they split the work between it and deterministic processing.

Wrap-up

• OpenAI Build Week was a challenge built around Codex and GPT-5.6, drawing about 47,000 participants from 186 countries with over 8,000 projects submitted
• Winners span 4 categories and 8 projects, with a $100,000 total prize pool — each category's 1st place gets DevDay access, a meeting with the Codex team, and a year of ChatGPT Pro
• Genres range from accessibility to medical triage, developer security, and educational content
• Every project pairs GPT-5.6 with deterministic processing rather than relying on it for everything

If you want to build something with Codex or GPT-5.6, or you're just curious how other developers are approaching app design, this is a great collection to draw inspiration from!