How to use Claude to remake SimTower for the browser
The author spent six months using Claude Code to rewrite SimTower for the browser, supplying decompiled references and gameplay memories, then validating behavior through extensive playtesting.
news.ycombinator.com
Suggested prompt
Suggested prompt written by Agents In Use, not the source author. Replace bracketed inputs and review tool permissions first.
Help recreate [game] for the browser using [decompilation], [source summary], and [gameplay notes]. Target WebAssembly, extract assets from a user-supplied original binary, and incorporate behavioral corrections from my playtesting.
Reach people exploring AI agents
See placements →Evidence at a glance
Sourced account- Source poster / publisher
- jonahss
- Original account
- Read the source ↗
- Published
- Source link checked
- Reported result
- The author reports deploying a browser-playable SimTower rewrite as WebAssembly after six months of work, with original assets extracted from a user-supplied binary.
- Human involvement
- The author did not read generated code, but spent hours playing the game and scrutinizing its behavior.
A linked account supports attribution. It does not mean we independently reproduced its result.
Report an issue with this case
Open correction emailOpens your email app with the case link included. Nothing is sent automatically.
How the agent did it
- The author supplied a Ghidra decompilation, an AI-generated source summary, and gameplay memories as references.
- Claude Code worked with the author on a complete SimTower rewrite.
- The author spent hours playtesting and correcting small behavioral details.
The author wanted to replay SimTower but found getting it running through emulation too difficult. They instead spent six months working with Claude Code on a complete rewrite for the browser. The reference material comprised a Ghidra decompilation of the binary, an AI-generated summary of the original source code, and the author's memories of playing the game.
The author describes the goal as faithfully replicating the original functionality and feel. They avoided creating new artwork. Instead, users supply an original binary, and code on the page parses it to extract the original assets. The rewritten game was deployed to the web as WebAssembly.
Oversight came through playing rather than reading the implementation. The author says they did not inspect a single line of Claude's code, but spent hours playing and questioning small behavioral details. Elevator behavior received particular attention, including time spent deciding whether its difficulty reflected the original game. The post shares a playable browser version, a writeup, and source code, but supplies no independent assessment of fidelity.
Who it's for: Developers exploring agent-assisted game recreation who can provide reference material and invest substantial time in behavioral playtesting.
Prompt
Suggested prompt (written by Agents In Use, not the author's)
Source: Hacker News · jonahss
Frequently asked questions
Where does the browser version get its artwork?
Users provide an original game binary. The author says the on-page code parses that binary and extracts the original assets.
How did the author check Claude's work?
The author reports hours of playtesting and detailed feedback about gameplay behavior, while explicitly saying they did not read Claude's generated code.
More details
Remaining steps
- The rewritten game was deployed to the web as WebAssembly.
About this account
Claude Code produced the implementation while the author supplied references and directed behavioral corrections through playtesting.
Platform: Hacker News. Attribution: https://news.ycombinator.com/item?id=49702448. Prompt: Suggested (written by Agents In Use, not the author's). Sourced by the Agents In Use pipeline.
Credit: jonahss on Hacker News
Source published
Added to Agents In Use
Source checked
Tools and topics
Claude Code · Ghidra · WebAssembly
Game recreation · WebAssembly · Playtesting
Editorial completeness: 100/100. Measures documented fields, not agent performance or verified outcomes.