I'm an engineer, but these days I solve the boring problems entirely by vibe coding. Somewhere along the way I felt I had lost the creativity and curiosity I used to have when writing code myself.
So I started thinking about how to make vibe coding fun again. codyssey turns a Claude Code session into a little adventure: while the agent does the work, you hear it as music, watch a pixel knight fight through it, and read it back as a quest log.
delexw
/
codyssey
Turn every Claude Code session into a little adventure: generative music that follows the agent's mood, a pixel knight who battles a monster for every edit and command, and a transcript retold in game style.
codyssey
Claude Code, played as an adventure: soundtrack, hero and quest log.
Important
miod's music only works on macOS.
I'm an engineer, but these days I solve the boring problems entirely by vibe coding. Somewhere along the way I felt I had lost the creativity and curiosity I used to have when writing code myself.
So I started thinking about how to make vibe coding fun again. codyssey turns a Claude Code session into a little adventure: while the agent does the work, you hear it as music, watch a pixel knight fight through it, and read it back as a quest log.
It's three mods in one plugin, and each one turns on and off on its own.
What it is
Three Claude Code mods in one plugin.
miod: the soundtrack
Generative music written on the spot while Claude works. No music files, no library, no AI model.
- …
Top comments (1)
Love the "vibe coding" framing — it captures that shift from syntax-wrangling to intent-expression perfectly. The game metaphor works because it exposes the agent's decision loop: state → action → reward → next state, which is exactly how we should be evaluating autonomous systems (evals as "levels", regressions as "boss fights").
Curious how you're handling the non-determinism in the reward signal. Are you using a fixed rubric, LLM-as-judge, or something hybrid? Also, did you hit the "reward hacking" wall where the agent optimizes for the metric instead of the actual goal? That's usually where the game analogy breaks down — or gets interesting (site: labagent .tech)