Replit

Free the models: Harness design at the frontierWhy Replit Agent lets the core loop pick subagent tier, effort, and specialists—and beats rigid routers on cost/score

Replit's AI team argues model routers are always weaker than the models they choose for. Their harness lets GPT-6 Astra decide effort and delegation; on DeepSWE and Terminal-Bench, Replit Agent is Pareto-efficient versus Astra alone and a sidekick architecture.

Blog post · AI Agents · AI · Engineering · Benchmarks