NVIDIA engineers ship 10x faster with Codex agent
Curated by the Inblix editorial team
NVIDIA’s internal team is using Codex — an AI coding agent powered by GPT‑5.5 — to speed up everything from building production apps to running ML research experiments. The tool runs on NVIDIA’s own GB200/GB300 infrastructure and handles long, autonomous sessions without losing context. Senior engineer Dennis Hannusch said Codex turned an internal platform from MVP to production-ready, something earlier models struggled with. Another team built a podcast recording app in hours instead of weeks, with Codex even testing the app autonomously. For researchers, Codex automates the whole loop: reading papers, writing ML experiment scripts, and running them remotely. AI researcher Shaunak Joshi highlighted GPT‑5.5’s creativity for tracing evidence across research papers and suggesting knowledge graphs. Results include a 10x speedup in research workflows and 40,000 NVIDIANs gaining access. Why it matters: This shows how cutting-edge AI isn’t just for chatbots — it’s becoming the default tool for shipping real, complex software inside one of the world’s most hardware-focused companies.
💡 Key Takeaways
- Codex on GPT‑5.5 handles longer, more autonomous coding sessions with high accuracy and context retention.
- It enabled building a production app from MVP in weeks, and a podcast recorder in hours — tasks that previously took much longer.
- Research teams achieved a 10x speed improvement by automating the full ML experiment lifecycle from idea to execution.
Keep reading: See related articles below for more coverage on this topic.
Get smarter about AI
The sharpest AI news, curated daily. Delivered free to your inbox.