His team runs agent data jobs on small local models with a human alerted to review
His team runs agentic data-enrichment and evaluation jobs on small local models that alert a human to review scenarios, at about 94 percent real-world tool-call success.
How it works
- The team runs agentic enrichment and evaluation jobs on small local models
- The jobs alert a human to review scenarios
- Real-world tool-call success runs at about 94 percent
Why it’s worth knowing
Small local models plus a human backstop is a working cost shape.
Where it’s from
Summarized from a public post on news.ycombinator.com. See the original ↗