How the money works.
These are the rules I price by. I publish them so you can hold me to them. The figures sit with the offers on the Services page.
Six rules I price by.
No figures on this page. Those sit with the offers they belong to. This is the machinery, published so you can hold me to it.
Fixed prices, always
An hourly rate charges you for my learning curve. It pays me to work slow. A fixed price puts the risk of a bad estimate on me. You know the number before anything starts. It holds even when a week goes badly.
You pay a milestone when it passes
Every milestone carries written pass/fail criteria. You approve them before I build. Half is due when the milestone starts. The balance is due when the work passes the criteria you signed. So "done" is a test, not my opinion. The work passes it or it does not.
Prices only move one direction
I cap how many builds I take at once, so every one gets done well. When the waitlist runs past a quarter, prices step up ten to fifteen per cent for new work. I publish that rule instead of negotiating it. So the cheapest day to start is always today. That is plain arithmetic and I would rather you saw it.
Bonuses instead of discounts
When the price is a problem I add something rather than take something off. A discount admits the first number was padded. Then every number after it looks padded too. An extra training session costs me something real. An extra quarter of monitoring does the same. Both cost you nothing in trust.
You can leave whenever you like
Cancel anytime with one email. Your code lives in your own GitHub org from day one. On your way out you take a full export of your telemetry history. You take your eval suites and a final report. You keep every asset the work produced. You give up one thing only: the service that keeps running. That means future reports and your live benchmark position. It also means deprecation watch and drift alarms.
"You don’t need this" is a real answer
Discovery can end with me telling you to skip the work. It can end with a process change beating a system. It can end with your spreadsheet being fine. Each one is a real result and you still get the written reasoning you paid for. Anyone who quotes a build before discovery is guessing. That includes me.
Open for projects
Find the work worth doing.
Tell me what runs slow or costs too much or keeps breaking. If AI is the wrong answer I say so.
