LLM benchmarks: what each bench actually measures
Capability evals and systems benches answer different questions. A plain map of LLM benchmarks for eng leads choosing what to measure before they pick a model or agent stack.
Read the articleBetter Technologies,
Better Results
Notes on the technology decisions behind the systems we build.
Decision model vs LLM: when agents should classify, route, and gate
When should an agent use a decision model vs an LLM? A practitioner look at TypeSafe Jev use cases for tool approval, routing, and typed Choice/Score outputs.
How QA moved in the AI era
Releases got faster and bugs got sneakier. How QA roles, tools, and techniques moved from classic automation to AI-assisted testing, eval harnesses, and the work we still keep human.
The verification tax is the real agent bill
Agentic coding made generation cheap. The bill moved to CI, browser gates, review tokens, and avoidable reruns. How we cut that tax without softening quality.
Which backend agents actually handle well
Nest, Laravel, Django, ASP.NET Core, and Spring Boot compared for coding agents. Not AI SDKs. Golden paths, types, CLI, magic, and what public benches actually measure.