Building Codex with Tibo Sottiaux
Tibo Sottiaux is one of the engineers who created Codex, and today, he heads up the Core Products & Platform org at OpenAI which also includes Codex. He’s also one of the most public faces of Codex due to his frequent – and generous – usage reset announcements, like this one yesterday. In this episode of the Pragmatic Engineer Podcast, Tibo and I discuss how Codex was built and continues to be iterated upon. We explore why the Codex CLI is written in Rust and was released as open source, how the harness and models have evolved, and why Codex supports models from multiple providers. Tibo also shares details about how the OpenAI team uses Codex throughout the software development lifecycle, including code reviews, maintenance, and system rearchitecture. We look into how AI is lowering the cost of changing code – and some interesting side effects of this – the merger of ChatGPT and Codex, and also how Tibo uses the tools in his own work. — *Brought to you by:* • turbopuffer – a vector and full-text search engine built on object storage. It’s fast, cheap, and extremely scalable https://turbopuffer.com/pragmatic • Antithesis – verify your system’s correctness without human review or traditional integration tests – and avoid bugs or outages. https://antithesis.com/pragmatic • Entire – Git hosting, rebuilt for the agentic era. Every agent session, prompt and tool calls: stored in your repo. http://entire.io/pragmatic — *The Pragmatic Engineer deepdives relevant for this episode:* • How Codex is built https://newsletter.pragmaticengineer.com/p/how-codex-is-built • How Claude Code is built https://newsletter.pragmaticengineer.com/p/how-claude-code-is-built • How Cursor was built https://newsletter.pragmaticengineer.com/p/cursor • What is "loop engineering?” https://newsletter.pragmaticengineer.com/p/what-is-loop-engineering • How Uber uses AI for development: inside look https://newsletter.pragmaticengineer.com/p/how-uber-uses-ai-for-development • Why Ramp built its own in-house coding agent, Inspect https://newsletter.pragmaticengineer.com/p/why-ramp-built-inspect • “I ship code I don’t read”: with Peter Steinberger, the creator of OpenClaw https://newsletter.pragmaticengineer.com/p/the-creator-of-clawd-i-ship-code — *Where to find Tibo Sottiaux:* • X: https://x.com/thsottiaux • LinkedIn: https://www.linkedin.com/in/thibault-sottiaux-27195366 — *In this episode, we cover:* 00:00 Intro 07:21 Working at Google 12:41 What drew Tibo to OpenAI 15:19 The early days of Codex 18:20 Why Codex was built in Rust 21:15 Why Codex is open source 25:50 Codex plays nice with other models: why? 32:09 How the harness works 36:44 Harness and model improvements 41:19 The SDLC behind Codex 46:39 Code reviews at Codex 52:09 Maintenance and architecture 56:43 How AI tools expand what engineers can do 1:02:30 The Merge: ChatGPT + Codex 1:07:16 How Tibo uses Codex and ChatGPT 1:10:44 Advice for engineers who want to work in AI — See the transcript and other references from the episode at https://newsletter.pragmaticengineer.com/podcast — Production and marketing by https://penname.co/.
Video Chapters
- 0:00 Building internal artificial intelligence chatbots a year before ChatGPT launched
- 2:10 Joining OpenAI to work on fast moving reasoning models today
- 5:15 Early childhood in a remote village sparked computer interest early
- 8:47 Applying mathematics and optimization techniques to real supply chain systems
- 13:03 Working on Google infrastructure and transitioning into DeepMind research roles
- 15:20 Joining OpenAI to build infrastructure and explore early code agents
- 18:18 Choosing Rust for the core agent to ensure maximum performance
- 22:15 Deciding to open source the harness and build active community
- 25:58 Allowing multiple models to power the harness without vendor lockin
- 30:20 Discussing how developer tool competition drives faster innovation choices today
- 35:00 Exploring local versus cloud environments for running advanced coding agents
- 39:00 Understanding the relationship between the harness and underlying language models
- 41:30 How new engineers get onboarded and use internal artificial intelligence
- 44:00 The software development lifecycle and verification processes for massive user bases
- 44:36 Shipping code quickly and safely to a billion users daily
- 46:32 How superhuman artificial intelligence models are changing traditional code reviews now
- 51:57 Automating routine software maintenance tasks to save valuable engineering time
- 56:19 Rearchitecting complex systems and scaling infrastructure with modern tooling support options
- 59:26 Discussing the long road ahead for important human problems together
- 1:03:48 Exploring the merge of local coding agents into cloud environments
- 1:07:52 Personal workflows and managing multiple projects with artificial intelligence agents
- 1:12:08 Advice for software engineers wanting to build with artificial intelligence
Original Output
0:00 Building internal artificial intelligence chatbots a year before ChatGPT launched 2:10 Joining OpenAI to work on fast moving reasoning models today 5:15 Early childhood in a remote village sparked computer interest early 8:47 Applying mathematics and optimization techniques to real supply chain systems 13:03 Working on Google infrastructure and transitioning into DeepMind research roles 15:20 Joining OpenAI to build infrastructure and explore early code agents 18:18 Choosing Rust for the core agent to ensure maximum performance 22:15 Deciding to open source the harness and build active community 25:58 Allowing multiple models to power the harness without vendor lockin 30:20 Discussing how developer tool competition drives faster innovation choices today 35:00 Exploring local versus cloud environments for running advanced coding agents 39:00 Understanding the relationship between the harness and underlying language models 41:30 How new engineers get onboarded and use internal artificial intelligence 44:00 The software development lifecycle and verification processes for massive user bases 44:36 Shipping code quickly and safely to a billion users daily 46:32 How superhuman artificial intelligence models are changing traditional code reviews now 51:57 Automating routine software maintenance tasks to save valuable engineering time 56:19 Rearchitecting complex systems and scaling infrastructure with modern tooling support options 59:26 Discussing the long road ahead for important human problems together 1:03:48 Exploring the merge of local coding agents into cloud environments 1:07:52 Personal workflows and managing multiple projects with artificial intelligence agents 1:12:08 Advice for software engineers wanting to build with artificial intelligence Timestamps by StampBot 🤖 (593-building-codex-with-tibo-sottiaux)
Unprocessed Timestamp Content
0:00 Building internal AI chatbots a year before ChatGPT launched 2:10 Joining OpenAI to work on fast moving reasoning models 5:15 Early childhood in a remote village sparked computer interest 8:47 Applying mathematics and optimization techniques to real supply chains 13:03 Working on Google infrastructure and transitioning into DeepMind research 15:20 Joining OpenAI to build infrastructure and explore early code agents 18:18 Choosing Rust for the core agent to ensure maximum performance 22:15 Deciding to open source the harness and build a community 25:58 Allowing multiple models to power the harness without vendor lockin 30:20 Discussing how developer tool competition drives faster innovation choices 35:00 Exploring local versus cloud environments for running coding agents 39:00 Understanding the relationship between the harness and underlying models 41:30 How new engineers get onboarded and use internal AI tools 44:00 The software development lifecycle and verification processes for massive user bases 44:36 Shipping code quickly and safely to a billion users daily 46:32 How superhuman AI models are changing traditional code reviews 51:57 Automating routine software maintenance tasks to save valuable engineering time 56:19 Rearchitecting complex systems and scaling infrastructure with modern tooling support 59:26 Discussing the long road ahead for important human problems 59:27 Adapting to the accelerated pace of software development and engineering 1:03:48 Exploring the merge of local coding agents into cloud 1:07:52 Personal workflows and managing multiple projects with ai agents 1:12:08 Advice for software engineers wanting to build with ai Timestamps by StampBot 🤖 (593-building-codex-with-tibo-sottiaux)