Stop Rationing Tokens: Let the Harness Pick the Model — Kimchi by Cast AI

Why watch

Cast AI engineers present Kimchi, an open-source coding harness that optimizes LLM costs by measuring cost per task instead of cost per token, and automatically selecting the best model for each task. The talk covers three months of results from 300 employees and demos three tools: Ferment for milestone-based autonomous coding, Teleport for persistent agent sessions, and Studio for team collaboration. Targeted at engineering leaders and developers building AI-powered development workflows.