tech
May 17, 2026
Exclusive: A Conversation with Tibo from Codex on What Your Company Has to Become When the Model Can Actually Do the Work
Between the launch of the new Codex and GPT-5.5 and now, something happened in my own house that has stayed with me more than any benchmark. My wife, who is not an engineer, built and shipped a working full-stack app. She is using GitHub for the first time. That is one anecdote, not a trend, and I am wary of overreading it. But it is the cleanest signal I have for what the April release actually did. The model can now carry the work, and the surface area of who can ship working software has widened far enough that the question of where human judgment lives inside a company stops being a developer question and starts being a leadership one.
TL;DR
- The capabilities of new AI models (Codex, GPT-5.5) mean they can now 'carry the work' and perform significant tasks.
- This advancement widens the scope of who can ship working software, moving beyond traditional engineers.
- The core question for companies is no longer about AI's capability but about the placement and nature of human judgment.
- This shift necessitates a change in leadership instincts, moving it from a developer problem to a strategic business problem.
- Companies that fail to build the necessary 'five layers' of human judgment around AI risk being over-restricted or under-restricted, leading to potential incidents.
- Successful integration will lead to companies that are eventually 'impossible to catch'.