How I work: in writing, per milestone
A written brief comes in, a scope goes back, and every milestone ends in something you can run. There are no calls, and at the end another engineer can pick the system up without me in the room.
An engagement, start to handover
A brief comes in
You writeWhat the system has to do, what exists already, when it has to be live. Rough is fine, and specific is faster.
A scope goes back
I write backThe job as I understand it, cut into milestones, with what each one delivers and what each one costs. In writing, so you can hold me to it.
Every milestone ends in something that runs
I build, you approveYou run it yourself and approve it before the next one starts. If I have misread the job, we find out after one milestone, not at the end.
Everything is yours
You keepThe code, the deploy path, the tests and the written record of every decision. Another engineer can pick it up without me in the room.
The terms, stated once
- Pricing
- Per project and per milestone, against a written scope. Never per hour.
- Calls
- None. Everything that would be a meeting happens in a written thread, including the difficult parts.
- Approval
- You run each milestone yourself and approve it before the next one starts.
- Ownership
- The code, the deploy path, the tests and the decision record are yours on delivery of each milestone.
- AI agents
- They carry the volume: the typing, the tests, the checks. I decide the architecture, read what ships and answer for it.
- Not offered
- Staff augmentation, standing meetings, hourly billing, open-ended retainers.
How a system gets built, in four stages
Same order, every project: one engineer, AI agents carrying the volume, secure defaults and production standards from the first commit. The order never changes, because finding a threat model problem at stage four costs the whole project.
Research and threat model
Map the problem space, the attack surface, and the constraints that are not negotiable. What "done" means is written down before any code exists, so a change of scope has to be a decision somebody makes.
- Requirements and scope lock
- Security threat modeling
- Architecture decision records
Architect and prototype
Data models, API contracts, auth flows, infrastructure topology. The parts most likely to be wrong get answered first, with throwaway prototypes.
- Database schema and RLS policies
- API contract design
- Infrastructure as code
Build and harden
Small increments, each one shippable. Every feature carries its own security review, performance budget and AI-assisted code analysis before it joins the trunk.
- AI-augmented development
- Security audit per feature
- Automated CI/CD pipeline
Deploy and iterate
Docker-first releases with a rollback that has been tested. What happens after launch is measured, and the next increment answers what the numbers said.
- Containerized deployment
- Production monitoring
- Continuous improvement
Ready to write it down?
Write what the system has to do, what exists already and when it has to be live. I read every brief myself and answer in writing, with a scope and a price for each milestone.
Or write to contact@guglielmo.io