DeepSeek V4.1 impresses early testers with role-based subagent prompting
teortaxesTex · x · 2026-09-13
A tester found DeepSeek V4.1 surprisingly capable at role-based subagent setups: their AGENTS.md instructs agent teams to assign subagents human roles well represented in pretraining data to harvest diverse perspectives and internal audits. The tester joked the run crashed their laptop twice — a first for DeepSeek.
Related event: DeepSeek V4.1 Test: Assigning Human Personas to Subagents(2 posts)→
More from coding & agent
- BountyScout bot hourly-scans GitHub bounties and floods maintainers with AI applications — gleech · 2026-09-13
- Cognition's SWE-2 impresses early user with solid coding, but lacks Codex-style computer use — CtrlAltDwayne · 2026-09-13
- Better AI 3D output costs more upfront but saves tokens and time vs. endless iteration — chaseleantj · 2026-09-13
- No prompt one-shots a polished 3D asset — human-judgment iteration took 17 more minutes — chaseleantj · 2026-09-13
- Workflow tip: generate a reference image first, then iterate your 3D output against it — chaseleantj · 2026-09-13
- Skill-based 3D generation: 68k to 153k tokens but far more realistic leaves — skills set expectations, not teach — chaseleantj · 2026-09-13