the difference is so little, you over-optimize for the wrong things
@steipete
-
Plan Mode Unnecessary When Using AI Agents Directly
By
–
I never use plan mode. The main reason this was added to codex is for claude-pilled people who struggle with changing their habits. just talk with your agent.
-
NemoClaw Plugin for OpenClaw: New AI Tool
By
–
NemoClaw is a plugin for OpenClaw. This vibed article makes not much sense.
-
Opus 4.6 vs GPT 5.4: Model Capabilities Comparison
By
–
That’s a model issue, not a harness one. You barking up the wrong tree. Try Opus 4.6 with adaptive thinking or (if you prefer most instruction following) GPT 5.4 high.
-
Security team implements stricter secure defaults
By
–
I addd some folks that help with the security advisory onslaught, and they went a bit too hard on secure defaults.
-
OpenAI Search Feature Now Carries Associated Costs
By
–
You always pay for it in one way or another. We just added search from OpenAI if you use that model.
-
Plugin System Enables Community Contribution While Maintaining Security
By
–
There’s such a great plugin system available now. Anyone can work on this. I’m still bogged down by keeping the lights on and keeping it secure.
-
Stopping Running Agents: Control and Execution Management
By
–
nah, stop works while an agent is running, you can always stop things that are running.
-
Rising User Expectations Challenge AI Development Progress
By
–
They are making great progress. Problem is that standards what people expect are raising with every release as well.