Alexander Embiricos
Product Lead, Codex at OpenAI
Product lead for Codex at OpenAI, the company's coding agent; previously founded a startup for five years and was a product manager at Dropbox, described by OpenAI's CPO as 'simply the best.'
Dimension Profile
Key Themes
Episode Summary
Alexander Embiricos shares what it's like building Codex at OpenAI, where radical speed and bottoms-up autonomy are enabled by extraordinary talent density. He discusses Codex's explosive 20x growth, how it enabled the Sora Android app to ship in just 18 days, why his team is now focused on making code review easier rather than just code generation, and his belief that the biggest bottleneck to AI progress is human bandwidth, not model capability.
Leadership Principles
- → At AI companies, empirical learning matters more than planning — be ready, fire, aim rather than ready, aim, fire
- → Even with no more model progress, we're way behind on product — there's so much more to build
- → The bottoms-up approach only works with extraordinary talent caliber — you can't just deploy it anywhere
Notable Quotes
"Codex is a bit like this really smart intern that refuses to read Slack, doesn't check Datadog unless you ask it to."
— On the current state and limitations of AI coding agents
"Even if we had no more progress with models today, we are way behind on product. There's so much more product to build."
— On the massive product opportunity ahead even with current AI capabilities
"The current underappreciated limiting factor is literally human typing speed or human multitasking speed."
— On the bottleneck to AI progress and AGI
Want to know how you compare?
Take the Assessment