Show HN: Claude Code takes forever, so I filled the waiting with Pokémon

Show HN on Hacker News introduced a side project that runs while you wait on Claude Code. Wild Pokemon appear in the Claude Code session as the agent works. The thread is at https://news.ycombinator.com/item?id=49144843 and, at the time captured, had 2 points and 0 comments.

The product mechanics are simple on purpose: Claude Code is the host environment; the wait is the trigger; Pokemon surface during that idle stretch rather than as a separate app. There is no published architecture, benchmark set, or integration surface beyond that loop. The pitch is latency theater — something visible and interruptible while the model is still thinking.

For engineers who live in agentic CLIs, long-running Claude Code turns are a real workflow cost. A lightweight overlay does not speed inference, but it makes the stall legible and less likely to push people into context-switching. Builders shipping coding agents can treat idle UX as part of the product, not only tokens-per-second and tool-call quality.

The post sits in the same HN conversation that also covers Claude, Codex, and Fable. Claude Code is the slow path being patched with play; Codex is the adjacent coding-agent stack people compare against; Fable is another name in that thread cluster. At 2 points and no comments yet, this is an early Show HN signal, not a crowded product category with public adoption metrics.

Watch whether the project stays a Claude Code-only gag or grows hooks for other agent CLIs, and whether a second HN wave or a real repo/docs link shows up so others can run the same wait-time Pokemon loop without reverse-engineering it from the title alone.