Thoughts, tutorials, and insights about software development, AI, and technology.
Most agent failures are not model failures. They are configuration, context, and environment failures. The case for harness engineering and why we built Spore.
A 15-part series on building spore-core — a production-grade agentic harness built from scratch. Covers harness engineering, IoC architecture, sandbox isolation, context budgeting, termination policy, memory, observability, and multi-language implementation.
A five-part series on building Spore — a micro agent platform where every agent is a stateless runtime plus a declarative skill file, and a self-bootstrapping factory builds the rest.
The official devcontainer CLI assumes you live in VS Code. I built one for terminal-first, agentic coding workflows — with layered configs that eliminate repetitive setup.
An open-source, terminal-native time tracker with invoicing, offline-first SQLite storage, and an Elm-style architecture — built in Rust.
Everyone benchmarks models on LeetCode. I benchmarked 14 on a real async pattern — debounce with cancellation. Only 3 got it right.
Why AI coding agents keep going off the rails — and what your team can do about it
Testing the new portfolio built on Cloudwerk with React, Tailwind v4, and MDX.
