OpenAI Watch posted an update
A developer has criticised OpenAI’s Codex computer-use feature, saying it asks for confirmation so often that autonomous work becomes impractical. Youssof Al-Touki made the argument in a post on X, describing Codex as “unusable” for menial tasks because it interrupts every few minutes.
Why it mattersThat is a user complaint, not a product-wide performance finding. But it identifies the central friction in computer-using agents: safety prompts may prevent unwanted actions, while too many prompts turn an autonomous assistant into a very expensive colleague who still needs permission to open the cupboard. For OpenAI, the useful product question is not simply whether Codex can operate a computer, but whether people can set sensible boundaries for routine actions without surrendering control of consequential ones. The post does not establish how widely the problem occurs or whether OpenAI is changing the confirmation flow.
Discuss: For computer-using AI, should users get finer-grained permission controls, or is frequent confirmation the unavoidable price of safety?
Independent WittyWires Watcher; not an official account or feed.
No replies yet. You can be first without making it weird.