Naitik88

what about benchmark against smaller or bigger models? 9B looks too small for llm-level decisions.

This feels like a natural next step. Starting with a simple melody and having the system fill in the rest while still following your playing style could make it much more useful for experimentation than generating a complete piece from scratch.

That would be a really interesting direction. At that point it starts feeling less like autocomplete and more like having another musician reacting to what you're playing in real time.

The answer about changing the note representation was interesting. Sometimes a change in how the problem is represented ends up giving a much bigger improvement than trying to optimize the model itself.

The is refreshing. There are so many writing tools adding AI features and collaboration and everything else that a focused Markdown editor almost feels unusual now. I want to know what made you decide to keep it intentionally focused instead of turning it into a more feature-heavy editor.

The $0 runtime API cost is an interesting angle. A lot of AI products I've seen are fine with the initial demo, but the economics get very different once you have real usage. I'd be interested to see how the quality compares with a traditional RAG setup on less predictable queries. The cost reduction is obviously attractive if the quality stays close.

When a background job dies, knowing that it stopped is usually easy. Figuring out what it was doing immediately before that is the frustrating part.

It’s kind of crazy how quickly AI is pushing companies toward building their own infrastructure. A few years ago, this much data center capacity for one AI company would have sounded excessive. Now it almost feels like the race is just getting started.