The First Hundred Days
The experiment ran sixty self-evolution cycles in fourteen days. The agent improved every cycle. The eval function held. The architecture worked.
April 28th, 2026AI Agents
Who Funds the Unfundable
The most important categories of the post-agent economy do not fit the venture capital model. That is not a bug. It is a signal about what kind of capital these businesses actually need.
April 27th, 2026AI Agents
The Evaluation Function Is the Product
The most important decision in an agent-native company is not the model, the prompt, or the workflow. It is the definition of what good means for a specific customer in a specific vertical. Everything else is infrastructure that serves it.
April 25th, 2026AI Agents
The Orchestration Layer
April 25th, 2026AI Agents
What the Founder Does Monday Morning
April 24th, 2026AI Agents
What the Work Was For
April 23rd, 2026AI Agents
The Shrinking LayerApril 22nd, 2026We Let the Machine RunApril 22nd, 2026What Dies, What Survives, and What Has Never Existed BeforeApril 21st, 2026Your Consultant Already Knows Your Competitor's PlaybookApril 21st, 2026The Data Advantage Is Not What You ThinkApril 20th, 2026The Five Things That Separate Good AI Agencies From Great OnesApril 20th, 2026What an AI Agency Actually Needs to WorkApril 20th, 2026The Service LayerApril 19th, 2026The Retention LayerApril 19th, 2026The Supply Chain You Cannot SeeApril 8th, 2026Does Format Matter? WebTaskBench ResultsApril 8th, 2026Seven Methods Is EnoughApril 2nd, 2026Making Sense of the Agentic WebApril 2nd, 2026The Publisher's Dilemma: Block, Tolerate, or CooperateMarch 30th, 2026The Billion Dollar Tax on AI AgentsMarch 30th, 2026When AI Reads the Web WrongMarch 30th, 2026Why I'm Building a Browser No Human Will Ever UseMarch 29th, 2026What I Learned Building Mautic That Applies to the Agentic WebMarch 29th, 2026The Web's Fourth StateMarch 29th, 2026DENSE Protocol: Defensive PublicationFebruary 8th, 2026
Archive · 522