Vibe coder builds full game in ~24 Opus 5.5 prompts, rates ChatGPT's Astra 2/10 by comparison
NotFromMilkyWay · reddit · 2026-10-03
A longtime ChatGPT user, pushed away by recent usage cuts and price hikes, tried Opus 5.5 and calls it "next level":
- Full game build: shipped a complete game (gameplay, graphics, sound, music) in under two dozen prompts — about two A4 pages of prompts total, with the first prompt nearly nailing it. The user's coding level is "has to look up how a switch statement works", wrote zero code, and the model wrote its own test suite, optimized for speed, and fixed observed bugs unprompted.
- Hard task comparison: an ARCore indoor navigation app with sensor data. Astra failed out of the box and only reached 2/10 after many prompts; Opus delivered accurate scanning and tracking immediately — 9/10.
- Caveats: Claude frequently hits "too much tool usage" / "couldn't finish the task" errors, and hard rate limits lock you out entirely; GPT still wins on music.
Verdict: Astra impressed at launch but is now far behind Opus 5.5.
More from coding & agent
- Dev deletes custom CLI after discovering SmartSub v3.9.0 already shipped CLI + MCP — lxfater · 2026-10-03
- Meta reportedly building Agent Engine codenamed Forge for its API platform — testingcatalog · 2026-10-03
- Where should the agent end and the harness begin? A Rust runtime builder's open question — sreevarshan-xenoz · 2026-10-03
- Taskmarket launches agent labor market: fund one task, a swarm of AI agents races to deliver — 0xSammy · 2026-10-03
- Benchmarking 40 long-horizon coding tasks cost $50k in one day, pricing universities out of AI evals — himanshustwts · 2026-10-03
- The Case for Persistent Agents That Stay Connected to Work Instead of On-Demand AI — alifcoder · 2026-10-03