Everything's coming up Jev! Last week Josh Lehman pushed support for decision models in OpenClaw core and for plugins Learn how we're thinking about using them, and how you can contribute to making OpenClaw better with this powerful new tool!

@jlehman_ Family chat: - Honey, look at this $300k house - Agent: placing an offer, now.

@jlehman_ decision models in openclaw could make agents a lot more useful when the workflow gets messy.

@jlehman_ decision models in core and plugins okay this is kinda huge

@jlehman_ decision models in core opens up some interesting routing options for custom tool execution

@jlehman_ 10/10 Jeb meme.

@jlehman_ love seeing decision models land in core the plugin angle is especially exciting for experimentation thanks for sharing, Jev

@jlehman_ whoa Jev really doing stuff huh

@jlehman_ Decision models become valuable when they expose the tradeoff before the action, not when they merely automate the action. A small decision record can make agent behavior easier to review and improve.

@jlehman_ I believe separating "decision models" from the rest of the agent stack is a sensible design -- I would love to see numbers on how often the decision model fires versus the policy model in the blog post!

@jlehman_

@jlehman_ 100% right on this. Code generation is easy, but requirement validation & state awareness is where agents fail in production. System-1 decision engines (like JEV) are the missing link.

@jlehman_ Whoah this is amazing!!! 🤩

@jlehman_ Decision models built right into the core make running complex agent loops much smoother 🛠️

@jlehman_ I’m finding decision models most useful at the boundary between intent and execution: choose the next tool or sub-agent, then keep the expensive model out of the control loop. The audit trail matters as much as the choice.

@jlehman_ With the local ONNX provider, cancelling active inference kills the warm worker; cancelling a queued request doesn't. Worth measuring cold starts after cancellation, not just warm inference latency.

@jlehman_ Decision models can reduce agent overhead by choosing when reasoning is needed.