CourionAI Newsletter
← All news
coding-agents 2 min read

A 2003 PC classic on your iPhone: AI ported Command & Conquer to iOS in a weekend

A Google DeepMind designer used Claude Code and Fable 5 to port Command & Conquer: Generals Zero Hour to native iOS — first playable build in about 40 minutes, full source on GitHub.

A retro tank and radar dish rising out of a smartphone screen under pixel-art clouds

Here’s a fun one that says a lot about where AI coding tools have landed. Ammaar Reshi, Lead Product and Design for Google AI Studio, spent part of his weekend porting “Command & Conquer: Generals Zero Hour” — a PC real-time strategy game from 2003 — to iPhone and iPad. He didn’t write the port himself. He directed Anthropic’s Claude Code, an AI tool that reads, writes, and runs code on your machine, running the new Fable 5 model.

The result is a real native port, not a streaming trick or an emulator. The game runs directly on Apple’s ARM64 chips, with campaign, skirmish, and the “Generals Challenge” mode all working via touch controls. Under the hood, the AI translated the game’s ancient DirectX 8 graphics code to Apple’s modern Metal API through several intermediate steps — the kind of tedious, error-prone work that usually eats weeks. Reshi says the first playable build took about 40 minutes, followed by “a few hours” of debugging. Over two days, he burned through his entire Claude Max subscription quota. The full source code is on GitHub, including an engineering log documenting every bug and fix. Game assets aren’t included — you need your own copy, about $5 on Steam — and on iPads the game can crash in long sessions due to high memory use.

What’s behind this? A year ago, “AI can code” mostly meant autocomplete and small scripts. What’s new here is scope: an agent held a two-decade-old codebase, a graphics API translation, and a new platform’s quirks in its head at once, and shipped something playable in an afternoon. There’s also a nice subplot — Reshi works for Google, and when asked why he used a competitor’s product, he answered that “you can love the AI space and respect the competition while still being fully focused on building.” Worth keeping expectations grounded, though: this was an experienced product person steering, debugging took real hours, and the memory crashes show the port is impressive rather than polished.

What this means for you: If you’ve never touched an AI coding tool, this is a readable glimpse of what “agentic coding” — AI that works through a big task in many steps, not one answer — can do in skilled hands. For the curious: the GitHub repo and its porting playbook are an unusually honest look at the process, bugs and all. For power users, the quota detail is the practical takeaway: ambitious multi-hour agent projects can eat a whole subscription’s allowance in a weekend. Budget accordingly.

Sources

Source: https://github.com/ammaarreshi/Generals-Mac-iOS-iPad

Next story

OpenAI's cofounder wants AI with 'almost no interface' — and admits why the last attempt failed

Greg Brockman says people shouldn't have to learn software anymore. His candid admission about ChatGPT plugins explains both the vision and why it's still far away.

A computer monitor dissolving into drifting geometric petals above a tidy desk, with a doorway of light behind