OpenAI should have split ChatGPT and Codex into two products, argues strategist, not built a super app

signulll · x · 2026-10-01

signulll argues OpenAI's roadmap is now largely driven by X, and offers an alternative: skip the super app, split into two surfaces.

The super-app approach has created massive technical and cognitive complexity—too many modes, tools, agents and surfaces (like the new dots) that even normal users can't parse.
ChatGPT should stay the consumer brand with simple, proactive personal agents (a daily feed, stuff that just works); Codex should be the enterprise agent surface, deeply integrated with work tools, browsers, terminals and company context.
Clear product boundaries would make launches easier and allow opinionated UIs for specific users, while 90% of underlying infrastructure/code stays shared—complexity pushed into product boundaries, not onto users.
He acknowledges the counterarguments: IPO pressure, needing Codex growth to fend off Anthropic, compute constraints.
His gripe: it defaults to work when he wants to chat, and chat when he wants to work.

More from Companies & People