@MatthewBerman@MichealColhoun Honestly, they’ve cloned so many of the open claw skills that my Claude in the CLI and my codex and the CLI really are my most reliable and trustworthy AI right now
Yup pro through Oauth just describes what it would do and never does. Built-in definition doesn’t have reasoning:true. So openclaw never sends the reasoning block, no tool calls, text only completions. Fix was adding a provider override in the openclaw.json. Under the models providers 5.4 reasoning true.
@AlexFinn@duskofoed Reliably is a stretch (perhaps relatively). Ive been pretty disappointed many times with opus claw. Wonder if hermes + 5.4 may be as good or better?
@bcherny Do not use the free credits!!!! They run out almost instantly and then they give you almost nothing on the refund!!! Its a total scam, never using anthropic again!!!!!
@tyleryust@openclaw It just isn’t clear. The new add on to the TOS says no auth in a bot— everyone interpreting that as no claw. But do we have confirmed bans? Can’t a lobster get a tweet clarification at least?