Agent swarm pentest: 10 DeepSeek v4.1 agents in one VM self-organize to find exploits
teortaxesTex · x · 2026-11-11
Developer nateberkopec shares a surprisingly effective automated security testing setup: have Claude spin up an isolated VM (capped CPU/RAM, egress only to an LLM gateway), run 10 DeepSeek v4.1 1017 Pi agents inside with a pi-intercom for self-organization, give them the app's source code plus a running copy, and instruct the swarm to report every reproducible exploit consistent with your threat model. He describes the results as "extremely effective" — a low-cost alternative to continuous pentesting.
Related event: Developer Runs 10 DeepSeek Agents in Isolated VM to Hunt Bugs(2 posts)→
More from coding & agent
- Agent platforms answer 'Can I build it without code?' but almost none answer 'Will it hold in production?' — Ayanle_O · 2026-11-11
- Rork's built-in Mobbin search surfaces 600K+ real app screens as design references, free plan included — rudrank · 2026-11-11
- swyx asks AEO CEO why his agents keep choosing Resend on their own — AccBalanced · 2026-11-11
- Matt Shumer's new video clarifies when to use Fable 5.2 vs Opus 5.5 — prasenx · 2026-11-11
- Rabbit insider: OS3 auto-responds to bug reports and drafts fixes in seconds — jesselyu · 2026-11-11
- Anthropic launches Claude Meetings: Claude joins Google Meet calls and does tasks live — No-Gazelle1365 · 2026-11-11