The Sam Altman Codex email habit, disclosed on a recent episode of David Senra’s podcast, has surfaced an uncomfortable contradiction at the heart of AI adoption: the person running one of the world’s most prominent AI companies admits he still manually scrolls through his inbox, even though the tool to automate that work is sitting right in front of him.
On the episode, released Sunday, Altman was candid. ‘I should not be scrolling mindlessly through my emails and trying to figure out which one is least painful for me to open and respond when I don’t want to be dealing with it,’ he said. Instead, he said, he ought to be handing that kind of routine work to OpenAI‘s Codex, including moving information between applications and working through to-do lists.
Codex is OpenAI’s software engineering agent and coding assistant, capable of performing tasks on a computer with a particular focus on helping programmers write, review, and test software. It launched in April 2025.
What the Sam Altman Codex email habit reveals about human psychology
Altman named the behaviour directly, calling it psychologically inconsistent. ‘And yet there’s something in my mind that is encoded that doing this kind of stuff is what it means to work and what it means to be productive,’ he said. The tension, as he described it, is not about capability (Codex can perform the task) but about a deeply embedded human association between familiar manual effort and a sense of accomplishment.
‘If you asked me, I would never say I like doing it that way,’ he added. ‘It makes no sense other than I must secretly like it or feel good about it.’ He suggested that better product design could make the transition to AI-assisted workflows more seamless, removing the psychological friction that keeps users in old habits.
The observation lands differently when set against what is actually happening inside OpenAI. According to OpenAI’s own account of how agents are transforming work, through August 2025 the average OpenAI worker spent less than 10% of their tokens on Codex. That figure has since shifted: every department at the company, including non-technical functions such as Legal and Recruiting, now uses Codex as their primary AI tool for work. The gap between early, patchy adoption and near-universal use closed faster internally than Altman’s own inbox behaviour might suggest.
Meanwhile, the OpenAI Developers Blog notes that GPT-5.2-Codex became the default choice for code generation, review, and repository-scale reasoning by mid-to-late 2025, signalling how quickly the underlying capability has matured even as habits catch up more slowly.
How other executives handle inbox overload
Altman is far from the only chief executive wrestling with email. A Business Insider analysis of 100 ‘Power Hours’ interviews found that leaders commonly set priorities before opening their inboxes, check messages during designated time blocks, keep replies short, automate or delegate triage, and enforce evening cutoffs.
Klaviyo CEO Andrew Bialecki has taken the AI-first route Altman described but has not yet adopted for himself. In a March interview with Business Insider, Bialecki said he has built scripts that send him a consolidated summary of key information from the previous 24 hours, bypassing the inbox altogether.
Siemens CEO Roland Busch approaches the problem differently. He receives around 200 emails a day but aims to keep his inbox below 100, telling Business Insider that most of his replies are simply ‘OK’ or ‘no’. For longer responses, he sometimes dictates while walking through the office.
Altman framed the broader moment plainly: people are, in his words, ‘straddling these two worlds’, still using computers in familiar ways (manually sorting through email) while experimenting with AI tools that can perform those same tasks. His comments come as OpenAI confidentially filed to go public in June but has delayed those plans to likely next year. Codex’s rapid spread through OpenAI’s own departments suggests the straddling period, at least inside the company, is already shortening.


