Articles by Grizzly Peak Software

You can build an agent in 100 lines. Proving it still works after you change the prompt is the part nobody writes about — and the reason most demos never ship. Here's how to build a real eval suite in an afternoon, starting with a JSON file.

Your real LLM bill is a multiple of the per-token price you compared. The five hidden costs that decide it, and the one number worth measuring.

The bottleneck isn't the model anymore. It's your attention. Why I stopped prompting AI agents and started engineering loops that run without me.

Highlights from Alaska's Claude Code Meetup-covering sub-agent workflows, the skills vs. sub-agents debate, CLI tips, and insights from Anthropic's developer relations team.