OpenAI Announces Build Week Winners: Eight Codex-Powered Projects

On August 26, OpenAIDevs announced the Build Week winners, showcasing eight innovative projects built on Codex across four categories: work & productivity, education, life apps, and developer tools. The winning entries consistently combine LLMs with deterministic logic, landing in real-world scenarios such as healthcare, language learning, accessibility, acoustic design, and MCP security.

Confirmed

• Echo Canvas (1st, developer tools) by Kevin Yang is a browser-based spatial acoustics workbench supporting 2.5D and 3D modes, using GPT-5.6 for constrained generation and validation.
• Sentinel (2nd, developer tools) by Malik Bashaar Javaid audits MCP servers for security risks via static analysis, GPT-5.6 review, and Docker sandbox probing, catching unsafe execution and hardcoded credentials.
• veTriage (1st, work & productivity) by Erin Downes is a PWA triage app for veterinary teams that turns clinical triage expertise into a deterministic workflow, already used at a real hospital front desk.
• Pulse (2nd, work & productivity) by Mohamed Mostafa tracks clinical state during cardiac arrest (rhythm, defibrillation counts, medication timing, pause durations) using a deterministic state machine rather than LLM-driven clinical decisions.
• Second Voice (1st, life apps) by Ravitez Dondeti helps dysarthric patients convert unclear speech into sentences for review and approval, combining GPT-5.6 reasoning with personal phrasebooks and context.
• AirBridge (2nd, life apps) by Adam Tarantino streams Windows system or per-app audio directly to one or more AirPlay receivers such as HomePods, with no virtual audio cables or temp files.
• Mechanica (1st, education) is an interactive digital museum that reconstructs ancient Chinese machines as 3D exhibits via physics simulation, explained by a hallucination-refusing AI guide.
• Dấu (2nd, education) is a Vietnamese tone practice tool using DSP pitch extraction and template matching for tone visualization, with local core scoring requiring no API key and LLM-based pronunciation coaching.

Why it matters

• The winners share a design pattern: LLMs handle interaction, explanation, and language understanding while critical decisions go to deterministic workflows or local algorithms, balancing capability and reliability in high-stakes domains.

2026-08-26 ~ 2026-08-26 · 10 related posts

• [source] OpenAI Announces Winners of Build Week Hackathon — OpenAIDevs · 2026-08-26
• Build Week Winner: Mechanica, an Interactive 3D Museum of Ancient Machines — OpenAIDevs · 2026-08-26
• Build Week Runner-up: Dấu, Visualizing Vietnamese Tones for Learners — OpenAIDevs · 2026-08-26
• [source] veTriage wins first: veterinary triage PWA in clinical use — OpenAIDevs · 2026-08-26
• Pulse assistant tracks clinical state during cardiac arrest — OpenAIDevs · 2026-08-26
• Second Voice aids dysarthria patients with zero-training AI — OpenAIDevs · 2026-08-26
• AirBridge streams Windows system audio to AirPlay speakers — OpenAIDevs · 2026-08-26
• [source] Echo Canvas wins first: browser-based spatial acoustics workbench — OpenAIDevs · 2026-08-26
• Sentinel tool detects MCP server security flaws via static analysis — OpenAIDevs · 2026-08-26
• OpenAI Build Week winners announced: Sentinel and more — OpenAIDevs · 2026-08-26