DeepSeek rolls out experimental Agent Teams feature, integrated with model training

teortaxesTex · x · 2026-09-10

DeepSeek users spotted a new experimental "Agent Teams" feature, with official messaging saying it is "deeply integrated with model training" and inviting early trials.

Observers noted DeepSeek already had subagents but this feels subtly different, likening it to Tachikoma-style swarm intelligence — and pointing out it took only about 9 hours from concept to rollout.

Original post →

More from coding & agent

coding & agent channel →