Let an agent refactor a 300k-line game decompilation with replay tests
CodeScene ran Claude Code on the Street Fighter III: 3rd Strike decompilation, about 300,000 lines. Frame-by-frame replays checked that every change kept the game behaving exactly the same.
Source
@_ninetyhe on X · 2026-10-18
Written in our own words. Credit to the original author. Link checked 2026-10-21.
What they did
CodeScene ran Claude Code on the Street Fighter III: 3rd Strike decompilation, about 300,000 lines. Frame-by-frame replays checked that every change kept the game behaving exactly the same.
Result (as claimed)
3 weeks, 2,903 commits and about $4k; code health went from 5.6 to 10.0.
Tools they used
Cost: Claude Code: from $17/mo.
Set it up
Pick a contained repo or task
Start on a branch or side project where mistakes are cheap.
Install and configure
Set up Claude Code, CodeScene. Read what it can run on your machine and limit permissions where possible.
Run it on a real task
Give it a well-scoped issue and let it work. Review the diff like any pull request.
Add guardrails
Keep tests and CI in the loop, and never let an agent push to main or touch production secrets unattended.
Copy the full setup prompt
Paste into your agent and it will walk you through the setup, asking before anything that sends, posts or spends.
You are helping me set up this AI use case end to end in Claude Code. Goal: Let an agent refactor a 300k-line game decompilation with replay tests What the original author did: CodeScene ran Claude Code on the Street Fighter III: 3rd Strike decompilation, about 300,000 lines. Frame-by-frame replays checked that every change kept the game behaving exactly the same. Reported result (as claimed by the author): 3 weeks, 2,903 commits and about $4k; code health went from 5.6 to 10.0. Source: https://x.com/_ninetyhe/status/2111753402111651932 Tools they used: Claude Code, CodeScene Category: Dev & ops Please: 1. Ask me 3 short questions about my situation (accounts I have, budget, what I want the output to look like). 2. Propose a step-by-step plan using these tools or cheaper alternatives, with the time and cost of each step. Use official pricing pages; say "check pricing" if unsure. 3. Then do the setup with me one step at a time, waiting for my OK before anything that sends messages, posts publicly, spends money or deletes data. 4. Finish with a checklist to run it again next time.
You are helping me set up this AI use case end to end in Codex. Goal: Let an agent refactor a 300k-line game decompilation with replay tests What the original author did: CodeScene ran Claude Code on the Street Fighter III: 3rd Strike decompilation, about 300,000 lines. Frame-by-frame replays checked that every change kept the game behaving exactly the same. Reported result (as claimed by the author): 3 weeks, 2,903 commits and about $4k; code health went from 5.6 to 10.0. Source: https://x.com/_ninetyhe/status/2111753402111651932 Tools they used: Claude Code, CodeScene Category: Dev & ops Please: 1. Ask me 3 short questions about my situation (accounts I have, budget, what I want the output to look like). 2. Propose a step-by-step plan using these tools or cheaper alternatives, with the time and cost of each step. Use official pricing pages; say "check pricing" if unsure. 3. Then do the setup with me one step at a time, waiting for my OK before anything that sends messages, posts publicly, spends money or deletes data. 4. Finish with a checklist to run it again next time.
You are helping me set up this AI use case end to end in Grok Bot. Goal: Let an agent refactor a 300k-line game decompilation with replay tests What the original author did: CodeScene ran Claude Code on the Street Fighter III: 3rd Strike decompilation, about 300,000 lines. Frame-by-frame replays checked that every change kept the game behaving exactly the same. Reported result (as claimed by the author): 3 weeks, 2,903 commits and about $4k; code health went from 5.6 to 10.0. Source: https://x.com/_ninetyhe/status/2111753402111651932 Tools they used: Claude Code, CodeScene Category: Dev & ops Please: 1. Ask me 3 short questions about my situation (accounts I have, budget, what I want the output to look like). 2. Propose a step-by-step plan using these tools or cheaper alternatives, with the time and cost of each step. Use official pricing pages; say "check pricing" if unsure. 3. Then do the setup with me one step at a time, waiting for my OK before anything that sends messages, posts publicly, spends money or deletes data. 4. Finish with a checklist to run it again next time.
The Daily 5
Get five new use cases tomorrow morning
Free. Real sources. Unsubscribe any time.
Guides for this use case
How to vibe code an app with Claude Code ยท How to vibe code a game with AI
Related use cases
Audit product copy line-by-line with a Claude Code skill
Built a Claude Code skill that tests each line of UI copy by asking what a user would misunderstand without it.
Send results straight from your coding agent to a teammate's agent
Instead of copying Codex output into Slack so a teammate can paste it into Claude Code, a developer built AgentDrop, an open-source tool that sends results, files and context directly from one agent to another.
Sync coding agent sessions and memory between laptops
A developer built mem-mastero to carry Claude Code, Codex and Copilot CLI sessions and memory across machines, using GitHub or Google Drive as the sync store.
See which routes and tests your agent's uncommitted change touches
He built LayerMap: it reads your uncommitted diff and lists the HTTP routes and existing tests that reach the changed code, so you know what an agent's edit might break.