Don’t know where to start?
Don’t know how you got this deep?
AI consulting for all levels.
I build with these tools every day — shipping a real product with them, and writing about what actually works when the repo is older than the model. Both of those questions above are ones I have had to answer for myself, more than once.
- Anthropic’s go-to-market team uses an MCP server I built, every day. Not a demo. A tool with users who would notice if it broke.
- DataRobot employee #36. Six years and nine months, from senior engineer to Director of Engineering — DevOps, release automation, and the internal tooling that keeps a company shipping while it triples in size.
- AI engineer and practitioner for over a decade. Since 2014, which was several hype cycles before this one.
- Founder and CEO of ItemMap. 3D reconstruction that runs off a phone camera, shipping on iOS, Android and the web. Built with the tools I would be advising you on.
- Four years at Crossbeam as a senior engineer. Data platform work on a product other companies depend on.
- Started out as a formulation chemist at Procter & Gamble. I came to software sideways, which turns out to be useful when the job is explaining it to people who did too.
Just starting
You have been told you should be using this. You would like a straight answer about where it helps, where it does not, and what it costs — before anyone commits a quarter to it.
Already deep
There are agents writing code nobody has read, a context file that has grown to a thousand lines, and a migration that is somehow both finished and not started. This is the more common call, and it is a fixable situation.
Making it stick
Turning a dev environment into something that takes a spec and returns a verified feature — encoding the business rules, the jargon, and what “done” means, so the agent stops guessing at the things only your team knows.
Tell me which of the two questions is yours. Either is a fine place to start.