Cut scope, not corners: an honest way to scope an MVP
How to scope an MVP without building something you have to throw away: the three features that prove the product, and the corners that never get cut.
Read the essay Essays
Written by the people who build it.
Technical essays on agentic AI, local LLM deployment, and shipping AI products. The methodology is the proof, so we write it down. No listicles.
Anatomy of a production LangGraph agent
What changes when a LangGraph agent has to run unattended: state design, durable checkpoints, human-in-the-loop interrupts, error recovery and evaluation.
Read the essayWhen to run your own model instead of calling an API
Data residency, cost at real scale, latency, the honest quality gap against frontier hosted models, and the operations burden nobody prices in.
Read the essay