One prompt
Each model gets the exact same self-contained mission and no staged follow-ups.
A real-world AI benchmark
Can an AI actually ship a game?
One prompt. One frontier model. One real iOS game—from blank folder to App Store Connect. The score is how often a human had to step in.
run_001.log
01researchdone
02designdone
03build + testdone
04store setupdone
AUTONOMY?FIRST RUN SOON
› ready_to_submit_
RESEARCH ✦ DESIGN ✦ BUILD ✦ TEST ✦ SHIP ✦ SCORE THE HELP
The question
Coding is the easy bit. Can it finish the job?
Most benchmarks stop when the code compiles. Ship a Game keeps going: market research, product taste, assets, simulator testing, screenshots, metadata, signing, pricing, and the last mile through App Store Connect.
The scoreboard
Games shipped by AI
Every finished run becomes a paid, playable iOS game—and an open account of what happened.
Rig ready · models warming up
The first run is coming soon.
The Mac is provisioned, the challenge is locked, and the intervention log is waiting. No results will appear here until a game genuinely reaches “Ready to Submit.”
The methodology
A fair fight on the same rig.
The same Mac, the same challenge, the same tools, and the same finish line. The model and its first-party coding harness are the recorded variables.
Its own cockpit
Claude uses Claude Code, GPT uses Codex, and Gemini uses Antigravity—the real product a user would choose.
The same loaded Mac
Xcode, native Swift, Godot, Unity, commercial-safe image and audio generation, CC0 assets, and App Store tooling.
A real finish line
The run only counts when the build reaches App Store Connect at “Ready to Submit.”
THE STANDARD RIG
Swift + XcodeGodot 4.7Unity 6Image generationAudio generationASC CLIThe headline metric
The intervention log is the score.
Every human nudge, fix, or rescue is recorded with its phase and severity. Known Apple account-holder steps are excluded. Everything else counts.
EXAMPLE SCOREONE-SHOT
100/100
autonomy
autonomy
A perfect run self-unblocks all the way to the finish. The published report keeps the full log—not just the flattering number.
01AutonomyHow much human help?
02Time to shipPrompt to ready-to-submit
03Run costModel + generation spend
04Game qualityIs it actually good?
Why the games are paid
A tiny ticket for a rather expensive experiment.
A game will typically cost around US$1.99. It is not a get-rich scheme; it is a small way to keep the runs going—and a chance for you to play the evidence.
Each run uses hours of frontier-model time, image and audio generation credits, a paid Apple Developer membership, and real operator time for provisioning, observation, and Apple’s required account steps.
Ask us anything about the experimentSHIP A GAMERUN RECEIPT
- Frontier model timehours
- Generated art + audiocredits
- Apple Developer Program$99/yr
- Human operatorreal time
YOUR PRICE≈ $1.99
Thanks for helping fund the next run. You also get the best seat in the house: judging the game yourself.
What we promise
Honest runs.
Playable proof.
No hidden hands.
✓Transparent intervention logsEvery meaningful human assist is part of the public story.
✓Commercial-safe assetsEvery shipped asset is generated with commercial rights or license-logged.
✓Real App Store gamesNo toy demos dressed up as benchmark wins.
Follow the experiment
The first model is at the starting line.
Want the first result—or have a model you think should take the challenge?
Keep me posted