Agents are eating the enterprise stack from the top down. These posts follow the consequences: which layer becomes the system of record, why seat-based pricing is the fault line, what happens to the firms that sell advice by the hour, and why the model is the smallest part of a working system.

1. Agent Swarms Are Eating the Stack Top-Down — Agents dissolve the application layer — but not the foundation underneath it.
2. The Substrate War — Does your data platform become your system of record, or the other way round?
3. Buy, Build, or Orchestrate — The old question retired. The new one is which layer you own and whether your price tracks value.
4. Service-as-Software — Selling a job done rather than software — and the four things that separate the winners.
5. The Frontier Labs Are in the Squeezed Middle — The labs aren’t exempt from their own thesis.
6. Who Advises the Disrupted? — If agents commoditise software priced by the seat, what happens to advice priced by the hour?
7. Agentic AI Is Not Just LLMs and Tokens — The harness is the system. The model is one component plugged into it.
8. A Product Playbook for Agentic AI — The frameworks I actually reach for when the spec is probabilistic.
9. Why Agentic AI Transformation Is Hard — Everyone treats it as a technology question. The hard parts are the ones you already know.
10. Give Your Coding Agents a Process — The productivity came from process, not a smarter model.
11. A Longer Leash — Self-verification is not verification — so the independent check matters more, not less.
12. Entity Resolution Is the Bottleneck — An agent that can’t tell two records apart can’t be trusted to act on either.

← All series