Developer reviews first 48 hours with OpenAI Dots
Onboarding is smooth, Codex thread orchestration is powerful and the calling feature works well, but it is slower than the demo, suffers from context bloat, often drops tasks midway, and does not auto-load the local machine's SKILLs and MCPs.
My [honest] first 48 hours using Dot: - Really great onboarding UX - Not nearly as fast as demoed - Native codex thread orchestration is really powerful - Unfortunately doesn’t auto-load the connected computer’s SKILLs and MCPs. Only uses ChatGPT plugins. However, it can spawn a codex thread on your computer which obviously does have those capabilities. Also, would be great to be able to connect multiple computers. - Super good for the first 12 hours - Suffers from context bloat - Unable to act as complete task manager, often forgetting or dropping work halfway through. Always sending reminders to follow up or recall specific instructions I gave it 20 prompts before. - Calling feature is still really awesome, I try to have it work on at least 1-3 tasks in the background at any given time. - Bonus: Spaces are very useful within the Codex Desktop App to do work all in on place. Will be interesting to see OpenAI compete with Notion lol. If you’re considering trying Dots, hope this helps! OpenAI is in a really good position to win consumer AI. Curious to hear other people’s experiences with agents like Muse, Instinct, or Grok Bot. Still really enjoy using my Hermes and owning my agent infra.
This post quotes another post: x.com/openaidevs/status/2105708732323909827