Latest field note
How We Vet an AI Product Before Launch
The practical product, security, reliability and operational checks AlanOps uses to decide whether an AI product is ready for customers.
Read the articleNOTE
Field notes
What I’m learning while building AI products.
Practical notes on AI products, CTO decisions, and the platforms that keep them reliable.
NOTE
Build, Buy or Wrap an Existing AI Product? A Founder’s Decision Framework
A practical framework for deciding which parts of an AI product should be custom, bought from a vendor or assembled around proven services.
Fully Hosted or Handed Over? Choosing the Right Operating Model
Compare managed hosting with an engineering handover, including ownership, support, security, costs and the exit plan you should demand.
What Does It Actually Cost to Build an AI Product?
The honest cost of an AI product depends less on screen count than on uncertainty, integrations, data, risk and the standard required at launch.
From AI Prototype to Production: The Work the Demo Doesn’t Show
A compelling AI demo proves a possibility. Production software must also be secure, observable, affordable and dependable when reality arrives.
Why We Don’t Quote Software From a PRD Alone
A PRD is a strong starting point, but a responsible software quote must resolve the assumptions that control scope, risk and delivery.
Is Your PRD Ready to Build? A 15-Question Readiness Test
Use these 15 questions to expose the product, technical and delivery assumptions that must be resolved before development starts.
Vetted Software: What It Means Now That AI Can Generate Code
AI has made code faster to produce, but dependable software still requires product judgement, verification and operational ownership.
What Two AI CTO Seats Teach You About Building Software Customers Actually Use
AMPD and CarBuddy AI sit at different points in the customer journey, but they keep teaching the same lesson: the hard part is not…
Yet Another Day in the Life: When Three Problems Pretend to Be One
Sometimes the universe decides that one production incident isn’t enough. It needs to stack them like Russian dolls, each one revealing another surprise when…
Working on something difficult?