Things worth returning to

The spec I build tooling against — worth re-reading as it evolves.

First stop for benchmarks when a paper lands on arXiv.

Prompt patterns and agent best practices, kept close.

One page that still teaches me new tricks every year.

The DX I measure other tooling against.

The paper that frames how I think about model size and data.

Local inference workhorse. Its design is a study in minimalism.

The documentation standard I try to match.

Pragmatic recipes for real-world LLM work.

Re-read before every big architecture decision.