FLIGHT 01Anthropic
Opus 5.5
AAAAAAAAPOLLO
A lunar landing mission with a very loud flight plan.
Play Opus 5.5 ← All projects 
Same prompt.
One-shot coding experiment
Lunar Scream.
Four models. One prompt. A very loud landing.
We asked ourselves "how can we make the most ridiculous test for GPT6-Sol and Opus 5.5?" So we came up with screaming. We threw GPT6-Astra and Fable 5.1 into the ring as well and let them rip.
You have the controls
Four takes. Try them all.
Use your voice for hilarity, or grab some friends for maximum ridiculosity. Each game has some surprises too.
FLIGHT 02Anthropic
Fable 5.1
SCREAM LANDER
Calibrate your microphone, find your pitch, and try to stick the landing.
Play Fable 5.1 FLIGHT 03OpenAI
GPT-6 Astra
Lunar Scream
Report to the flight department. Your voice is the propulsion system.
Play GPT-6 Astra FLIGHT 04OpenAI
GPT-6 Sol
AAAAA! / Moon Panic
Scream, steer, and see how gently you can arrive on the moon.
Play GPT-6 SolMicrophone permission is requested inside each game. Voice input is processed in your browser. For Fable and Opus, bring a keyboard for the full controls.
The rules were simple
Same prompt.
No second attempt.
This experiment isn't just to be absurd - it's to see what happens when you give a model something that goes well beyond their training data (and good sense). Each model got the same prompt, no clarifications or follow-up fixes. Everything was done in one shot.
The games retain their original gameplay, quirks, and bugs. There is no scorecard here. Play them and make your own call.
The exact promptBuild a scream-powered lunar lander as a playable browser game. Make it absurd and fun. Do not ask me any questions. Make your own decisions about what that means and how it should work, and take the project through to a finished, runnable game. You get one shot, with no follow-up feedback or requests for fixes. When you’re finished, tell me how to launch and play it.