Reading

Things You Should Never Do, Part I Joel Spolsky, 2000 · The canonical argument against rewriting a product from scratch, written after Netscape tried it
Strangler Fig Application Martin Fowler · How to replace a legacy system incrementally instead, moving behaviour across piece by piece while it keeps running
Opportunity Solution Trees Teresa Torres, Product Talk · Mapping customer needs against a business outcome, so discovery stays connected to what the company needs
Building Effective AI Agents Anthropic · The difference between a workflow and an agent, and the handful of patterns that actually hold up in production
AI Product Management, Two Years In Marty Cagan, SVPG · Why generative AI makes the product role more essential and harder at the same time
Which AI models are people spending money on? Peter Walker, OpenRouter · Model spend broken down by task across 29 categories, from a month of routing data
A basic recap of the AI story in 2026 Peter Walker, OpenRouter · Token demand, model diversity, and the collapse of open-weight pricing