Podcast Episode
The candid admission came just as OpenAI launched Codex as a standalone macOS application, designed for what developers call "vibe coding" - building software through conversational prompts rather than traditional programming. The tool allows developers to run multiple AI agents simultaneously, each working on different aspects of a project in organised threads.
Despite the competition, Codex has shown impressive adoption, with over one million developers using it in the past month. Usage doubled following the December rollout of GPT-five-point-two-Codex, OpenAI's most powerful coding model. Internally, Altman describes it as "the most loved internal product we've ever had."
This highlights a crucial aspect of human-AI collaboration: whilst AI tools can generate numerous possibilities and work tirelessly without losing motivation, human judgement remains essential for evaluating quality and relevance. The future may not be about AI replacing human creativity, but rather amplifying it through partnership.
OpenAI CEO Sam Altman Admits Feeling 'Useless' After AI Outperformed His Ideas
February 3, 2026
Audio archived. Episodes older than 60 days are removed to save server storage. Story details remain below.
OpenAI's Sam Altman made a candid admission that his company's new Codex coding app generated better feature ideas than his own, sparking discussion about AI's creative capabilities. The revelation came alongside the launch of Codex as a standalone Mac application, entering a competitive market dominated by rivals like Anthropic's Claude Code.
When the Creator Gets Outperformed
In a rare moment of vulnerability, OpenAI CEO Sam Altman revealed that his company's latest AI coding tool, Codex, came up with better ideas than he could. After building an app with Codex and asking it for feature suggestions, Altman admitted on social media that the AI's recommendations were superior to his own thoughts, leaving him feeling "a little useless."The candid admission came just as OpenAI launched Codex as a standalone macOS application, designed for what developers call "vibe coding" - building software through conversational prompts rather than traditional programming. The tool allows developers to run multiple AI agents simultaneously, each working on different aspects of a project in organised threads.
Fierce Competition in AI Development Tools
OpenAI's Codex enters an increasingly crowded market where it faces stiff competition, particularly from Anthropic's Claude Code, which reportedly reached one billion dollars in annualised revenue within six months. Major tech companies including Uber, Netflix, Salesforce, and Cisco have adopted Claude Code, whilst OpenAI counts various startups amongst its enterprise users.Despite the competition, Codex has shown impressive adoption, with over one million developers using it in the past month. Usage doubled following the December rollout of GPT-five-point-two-Codex, OpenAI's most powerful coding model. Internally, Altman describes it as "the most loved internal product we've ever had."
The Human Element Still Matters
Whilst Altman's admission sparked discussion about AI potentially replacing human creativity, some observers noted an important nuance. As one LinkedIn commenter pointed out, Altman only recognised the AI's ideas were better because he possessed the domain expertise to judge them. The AI didn't flag its own best suggestions - the human did.This highlights a crucial aspect of human-AI collaboration: whilst AI tools can generate numerous possibilities and work tirelessly without losing motivation, human judgement remains essential for evaluating quality and relevance. The future may not be about AI replacing human creativity, but rather amplifying it through partnership.
Published February 3, 2026 at 6:23pm
