From the source
Which tools are relevant to this request?
Which messages should survive compaction?
Is somebody actually talking to the agent, or should it stay quiet?
We can ask a language model to make those decisions, but doing so adds time and cost to the work the user actually wanted done.
As agents take on more of these small judgments, those extra calls start to matter.
Last week, Jev made a lot of developers rethink that tradeoff.
Today, Josh Lehman, an OpenClaw maintainer, shares why decision models caught his attention, how we’re bringing them into OpenClaw, and why some of the most interesting work is going to come from the community.
More than another model to chat with The magic is when you take Jev and embed it within normal deterministic code.
Josh’s first experiment was the obvious one: build a plugin that gave his agent a tool to call Jev.
It worked, but it left him with a slow language model thinking about when to call a fast API.
The more interesting possibility was to call it directly from application code.
…
