Claude Opus 5 built a functional Call of Duty clone. That is the lead. Anthropic's latest flagship model wrote playable game code from a prompt, and the Limitless podcast walked through exactly how it happened, what broke, and what had to be fixed.
The episode is not just a demo reel. It covers the agent stack behind the build, references Ledger's developer infrastructure for on-chain agent deployments, and gets into the practical limits of what Opus 5 can sustain across a long coding session. The specific failures are where the value is.
If you care about where LLM coding capability actually sits in 2025, this is a concrete data point. Watch it for the breakdown of the prompting workflow and the moment the model hits a wall, not just the finished product.
[WATCH ON YOUTUBE →]