I spent the weekend with Bonsai-27B, and I think I finally understand what a "27B that runs on one...
Bonsai-27B on a Single 3090: What Works and What Doesn't
Related
Why Athena/Iceberg Tends to Make Code the Spec
Whether 'code is the spec' holds is ~90% discipline, ~10% engine. Why Athena/Iceberg earns that 10% by externalizing state that humans and agents can reach — across OLTP and OLAP.
Your incident postmortems aren't investigations; they're fan fiction.
A look at why incident postmortems fail and how we can use MCP to enforce SRE-grade rigor.
Your A/B eval is paired. Your stat test probably isn't.
Two prompts on one eval set are paired data. The two-proportion SE said 'collect more'; McNemar said it was already decided.