Articles I return to.
- Observability: A 3-Year Retrospective — The New Stack — What observability actually is, why you want high-cardinality data, and why events beat metrics and traces.
- Use One Big Server — Speculative Branches — One beefy machine handles most workloads; at the time of writing, a single m5d.metal does 500k req/s to nginx and 70k TPS in Postgres. Vertical scaling beats microservice complexity for the vast majority of services.
- When Good Software Goes Bad: The Surprising Durability of an Ephemeral Technology — Nathan Ensmenger, The Maintainers
- Things You’re Allowed To Do — Milan Cvitkovic — A running list of things you can do but assumed you weren’t allowed to — hire a researcher, directly pitch yourself for a job, negotiate any bill, just ask for the crazy thing.
- After Automation — Dan Shipper, Every — Why AI progress creates more human work, not less. Benchmarks measure execution, but value lives in how AI gets embedded; as cheap competence floods in and creates sameness, humans stay structurally ahead by setting the next frame.