Curated GitHub Repositories

Sometimes you only need to read a file without opening the editor. bat adds useful reading context to that terminal moment, especially for source and configuration files.

You often remember part of a command or filename, just not the exact string. fzf is a useful way to make that partial memory work in a terminal workflow.

Local model inference gets interesting when you need control over the runtime. llama.cpp is a serious option if you are willing to match the model, build and hardware yourself.

Read nanoGPT as a compact reference for GPT training. The repository is explicitly deprecated, so check the author's newer nanochat project before starting fresh work.

For a first local-model experiment, reducing setup is useful. Ollama gives you a straightforward way to pull a supported model, run it and try it from an application.

A product event is more useful when you know the question behind it. PostHog is worth exploring if your team wants to connect instrumentation with actual product decisions.

Being able to change the component code is the appeal here. shadcn/ui is worth considering if you want a starting point for your own interface system and are prepared to maintain it.

The source is useful if you need to understand or operate the backend yourself. Go in with a plan for the services around Postgres as well as the database.

An infinite canvas has a lot of behavior hiding behind a simple screen. tldraw is worth considering when that canvas is part of your product, provided its production license fits your plans.

If your Python setup has become a collection of separate tools, uv is worth a small trial. Start with one project and check how its lockfile and commands fit your workflow.

If you want to understand what the model is doing during training, PyTorch is a sensible place to work. Be ready to own the data and compute decisions too.