How deleted vacation photos led to an open-source tool — and how building it revealed a structural problem that hits every team moving from experimenting with coding agents to using them in production at scale.
Why I built an agile AI team as a Claude Code plugin, how the SPARK loop works, and what happened when I ran a real feature through it.
A practical way to categorize modern AI systems beyond LLMs, from perception and simulation to action in the physical world.
A practical comparison of Qwen 3.6 and Claude 4.x on benchmark performance, token economics, and model selection strategy.
Engineering lessons from production-like long-form LLM pipelines with bounded retrieval, segmentation, and strong observability.
Why moving from manual prompt juggling to agentic end-to-end workflows significantly improved my development efficiency.
