One task, one thread, so bad corrections die with it
Give every task its own Codex thread so bad corrections die with it. Codex user fighting context pollution.
what to steal
Start a fresh thread per task; carry over only the memory you mean to.
Our takeaway from the source, not a quote.
niv-one-codex-thread-per-task.toml
# Give every task its own Codex thread so bad corrections die with it # source: https://x.com/niv_ai/status/2110222179749265499 [rig] who = "Codex user fighting context pollution" author = "Niv Asayag (@niv_ai)" date = "2026-10-14" harness = ["Codex"] models = [] [keep] - Each task is its own Codex thread - Only the worker's memory carries over [cut] - Long shared threads
| thing | status | why (per source) |
|---|---|---|
| Long shared threads | avoided | A messy correction dies with that task |
Source: https://x.com/niv_ai/status/2110222179749265499 · checked 2026-10-14
// similar rigs
~/rigs/paulund-herdr-cli.toml2026-10-05
Three desktop apps out, one terminal multiplexer in
steal → Use one session view for all CLIs instead of one app per harness.
CodexClaude CodeOpenCode Codex / Claude Code /
by Paulund
~/rigs/hboon-one-config-three-agent….toml2026-06-27
One config folder, symlinked into three agents
steal → Keep one AGENTS.md and skills folder; symlink it into every harness.
DroidClaude CodeCodex Per-harness config cop
by Hwee-Boon Yar
~/rigs/steipete-codex-grid.tomlundated
3–8 Codex agents, one folder, one 800-line AGENTS.md
steal → Have parallel agents make small atomic commits in one folder; swap MCPs for CLIs.
Codex gpt-5-codex Worktrees/PR-per-chang
by Peter Steinberger