Prompt & Done · FLAPPY pack
Same prompt, three AIs
Claude Opus 5.5, GPT-6.1 Sol and Qwen 3.5 (free, running on a laptop) got the same one-line prompt. Here are all three games exactly as they wrote them, the prompt, and all three rounds.
1 Play all three games
Round 1, unedited. Claude's and GPT's open in demo mode: the bird flies itself until you press a key. Qwen's opens on a start screen: press START DEMO MODE and watch.
2 The exact prompt
Paste it into Claude Code, Codex or any AI that writes code, then open the file in a browser.
3 All three rounds
| Round | Claude Opus 5.5 | GPT-6.1 Sol | Qwen 3.5 (local) |
|---|---|---|---|
| 1 | 1:18 · 26¢ ✓ | 8:33 · 37¢ ✓ | 7:37 · $0 ✗ bird lasts 0.3 s |
| 2 | 1:34 · 41¢ ✓ | 7:51 · 33¢ ✓ | 3:03 · $0 ✗ no file |
| 3 | 1:33 · 43¢ ✓ | 5:14 · 24¢ ✓ | 7:47 · $0 ✗ crashes on load |
Claude finished first every round; GPT was cheaper in rounds 2 and 3; the free model never produced a working game. Claude Code 2.1.296 and Codex CLI 0.162.1 (effort high); Qwen3.5-9B (Q4, Apache-2.0) in Ollama 0.35.0 via Codex, on an M4 Pro MacBook. Tested 11 Oct 2026. Costs are API-equivalent at list prices (both paid tools ran on subscriptions): OpenAI pricing · Qwen3.5-9B.