Decompile a full game with a long-running agent swarm

Goal: Run a multi-month decompilation with worker and reviewer agents.
Setup: agents used in source: Claude, Codex; tools/MCPs stated: Discord.
Steps:
1. Start with 3 worker agents decompiling and committing, plus 1 reviewer that coordinates and flags bugs; communicate via Discord.
2. Trigger compaction earlier (they cut the threshold from 90% to 42% context fill) to save tokens.
3. Stop the reviewer trusting workers’ comments: verify deviations against the original independently.
4. Build a verification harness, then scale workers and let them run autonomously.
Rules: ask me before anything irreversible (sending, posting, paying, deleting); don’t invent data, results or tools I haven’t given you; tell me what you couldn’t verify.
Based on: momo5502 — https://momo5502.com/posts/2026-10-09-game-decompilation/

Steps are from the source. Includes a “Based on” credit line.