LatestOct 11, 2026
Why few projects and one owner
Against hours-only and body-shop throughput. Ownership through to live, one interlocutor, handoff when it makes sense.
ReadNotes on product, AI automation, and technical decisions.
No editorial gloss.
Oct 8, 2026
Security feels boring, until you stop sleeping
Validating an idea does not secure your data. What belongs in live security scope, and what is not "phase two".
Oct 5, 2026
If the scope does not hold, we say so early
We are not the studio that always says yes. Examples of no, the cost of an easy yes, and how we work scope with founders.
Oct 2, 2026
An MVP in production is not a prototype
Demo, Figma, and staging are not live with users and data. What separates an MVP in production from a prototype, and which signal matters.
Sep 29, 2026
Smart automation or blind replacement: what separates growth from regression
Teams that get value from AI redesign roles instead of cutting headcount. Why automating to specialise beats automating to shrink payroll.
Sep 22, 2026
When the AI system stops working: why nobody knows what to do anymore
Handing a process to AI without keeping internal competence is a bet you pay at failure. What happens when it breaks, and how not to be left unable to act.
Sep 20, 2026
Local AI coding on a 64GB MacBook: cut costs without breaking Docker and IDE
Cheap cloud plus local models on a 64GB MacBook with Docker and IDE open. Cursor blocks BYOK in Agent Mode; Zed, LM Studio, Bonsai: what held in the 14 Sep 2026 session.
Sep 17, 2026
TypeSafe Jev: first impressions of typed AI decisions in production
First impressions of Jev (TypeSafe): classify, score, and yes/no with thresholds. LLM extractor + Jev supervisor. Vision, PDF, memory: still no.
Sep 15, 2026
Automation yes, not instead of the product
Automate after a product that holds. When it is an escape from validation, and how ownership stays on live.
Sep 8, 2026
AI does not help you sleep better: every extra layer you pay for in production
AI does not buy quiet nights: every extra layer is paid in production. How complexity lands in maintenance, and what to keep off money and data paths.