DeepSeek says it's betting on scaling across agent harnesses, pledging per-generation support

teortaxesTex · x · 2026-11-02

A DeepSeek researcher says the team, inspired by conversations with agent-tooling builders badlogic and mitsuhiko, believes in scaling across agent harnesses and will support pidot among major harnesses with each model generation. badlogic adds that DeepSeek is one of the few teams innovating in the harness space, and that Earendil shares the same direction: a single long-lived entity, ad-hoc freeform (possibly recursive) task delegation, and cross-system event ingestion and triggers.

Original post →

More from coding & agent

coding & agent channel →