A dedicated repo where research agents get the open internet

Fire off code-research tasks to async agents in a dedicated repo. Running 2–3 code research projects a day.

Share on 𝕏 ↗ Compare with… Read the source ↗

what to steal

Give risky, network-hungry research its own repo instead of loosening your main one.

Our takeaway from the source, not a quote.

Drawn from this card's sourced fields. Amber = older model named by the source. Rose = cut or replaced.

willison-async-research-repo.toml

# Fire off code-research tasks to async agents in a dedicated repo
# source: https://simonwillison.net/2025/Nov/6/async-code-research/

[rig]
who     = "Running 2–3 code research projects a day"
author  = "Simon Willison"
date    = "2025-11-06"

harness = ["Codex", "Claude Code"]
models  = []

[keep]
  - Dedicated research repos (one public simonw/research, one private)
  - Full network access in those repos
  - Short AGENTS.md with tips
  - GitHub Workflow auto-summarises projects into README
[cut]
  - Network lockdown

thing	status	why (per source)
Network lockdown	removed (research repo only)	Repo holds nothing sensitive, so prompt-injection exfiltration risk is low

Source: https://simonwillison.net/2025/Nov/6/async-code-research/ · checked 2026-10-14

Claude manages, Codex grinds

steal → Use one model to plan architecture and another to churn through build steps.

Claude CodeCodex Claude Fable as manageGPT-6 Astra on Codex

by @Merlin412188240

~/rigs/stefancrncic-t3-three-machin….toml𝕏 · 2026-10-13

Five harnesses, three machines, one orchestrator

steal → Install several harnesses on worker machines and drive them from one place.

Claude CodeCodexCursor

by Stefan (@stefancrncic)

~/rigs/mikeholp-route-claude-codex.toml2026-10-11

Route by task type and make each agent review the other

steal → Have the agent that didn't write the change review it.

Claude CodeCodexCursor Claude Opus 5.5Claude Sonnet 5.5 git push / commit in C

by Mike Holp