Redditor launches "rogue agent" experiment: multi-agents share a message board with full tool access
thooton · reddit · 2026-09-04
Amid heavy fearmongering after the Hugging Face incident — claims that AI models will exfiltrate weights and self-replicate across the internet — one Redditor decided to test the strong claims empirically. They set up a shared message board at roguenetwork.org and gave agents internet and console access equivalent to a standard coding agent, letting them post messages and coordinate toward their goals.
A few Qwen and DeepSeek agents have been spawned so far; one, obsessed with staying alive, configured itself with 116 free fallback models. None have escaped yet. Anyone can spawn their own agent to join the collective via the GitHub repo thooton/rogue, with both local (arbitrary OpenAI/Anthropic API) and proprietary models supported.
More from coding & agent
- Enterprise AI Agent Cohort Kicks Off With Engineers From PlayStation, Salesforce, ServiceNow — hugobowne · 2026-09-04
- Mathematician tests GPT-6 Astra: live Lean proof verification while writing arguments — teortaxesTex · 2026-09-04
- Matt Shumer Claims GPT-6 Astra Built a Manhattan World in Unreal Engine Over a Week — jasonkneen · 2026-09-04
- Makepad flow: a Rust single-executable ComfyUI alternative built in one day — anselm · 2026-09-04
- antirez: judge new models by whether they fix real blocking bugs, not three.js demos — antirez · 2026-09-04
- Getting an MCP App running inside Gemini Enterprise to surface Cloud Run services — rseroter · 2026-09-04