Beliefs in detail
Each of these pieces develops one idea behind our practice. They are specific enough to question, and useful only insofar as they help us make better decisions.
- 03.01 Taste as Infrastructure AI only differentiates when it carries a point of view. Encode judgment so teams can reuse it in tools, workflows, and decisions. Published
- 03.02 The Ratio Shift Human attention does not scale. More AI will move into background workflows, so the judgment inside those workflows has to be explicit. Published
- 03.03 Software Inversion AI moved the hard part from building software to knowing what should be built, how it should behave, and where it should fail safely. Published
- 03.04 Niche-Making Economics Collapsed build cost makes smaller, more opinionated companies more viable. You can be exactly right for a specific customer. Published
- 03.05 Factory Thinking Buyable features depreciate. A company that builds the factory can keep producing better tools, workflows, and customer experiences. Published
- 03.06 Capabilities, Not Systems Start with small useful capabilities that amplify people in existing workflows before committing to large systems. Published
- 03.07 Useful, Not Done Useful is a better milestone than done. Ship small improvements, learn from real use, and compound. Published
- 03.08 Friction Reduction The north star for AI work is lower friction - less searching, rework, handoff drag, and administrative load. Published
- 03.09 Capability Compounding Early wins should create context, tools, and confidence that make the next win cheaper. Published
- 03.10 Recursive Tooling We build our own tools, then use them to help customers build theirs. The method is visible in the work. Published
- 03.11 The New Consulting Paradigm The valuable partner is a senior builder who installs capability, not a deck consultant who leaves recommendations behind. Published
- 03.12 AI Pipeline Factory Chat is only one interface. Durable value often comes from structured workflows that extract, classify, route, synthesize, and trigger work. Published
- 03.13 Primitives Over Platforms When AI lowers the cost of engineering, durable cloud primitives often beat rented abstraction layers. Published
- 03.14 Exploration vs Execution Discovery and delivery need different modes. Know when you are learning and when you are hardening. Published
- 03.15 Collaboration Is Contribution What becomes easier, clearer, or more possible because we work together? That includes helping someone else become able to contribute. Published
- 03.16 The Happy Mascot Problem Office theater protects itself by turning the subject of a system into the system's cheerful spokesperson, but the positive aim is shared dignity. Published