Skip to main content

How a build runs, from the scope document to the day you own it

This page is for the person who has to defend the decision: an operations manager signing off on a sizable build, or a partner deciding whether to deliver one alongside me. It says how the work is run, in the order you will meet it.

The short version. One senior builder makes the decisions, and a crew of AI agents that I direct and check does the heavy lifting. I am the one who answers, and nothing is called done on a single pass.

Updated October 2026

From the scope document to the day you own it

  1. 01

    Scope

    A scope document with a fixed price, written before any work begins.

  2. 02

    Build

    Working software in steps you can see, while your feedback can still change it.

  3. 03

    Staging copy

    By default, built and shown on a copy, not on the system your team is using.

  4. 04

    Second check

    By default, a separate read-only pass checks the release. The pass that checks is not the pass that built.

  5. 05

    You acceptyour call

    You say yes on the copy. Until then, your live system is the one you already had.

  6. 06

    Handoff

    It goes live with a backup and a written record of what changed. The code, the domain and the data are yours.

Anything the check finds goes back to the build, and is checked again.

Scope

Every build starts with a conversation about what is not working, in your words. Then I write a scope document with fixed pricing before any work begins. It says what you are paying for and when you will have it.

The number in that document does not change unless the scope does. If the scope changes, we talk about it before I bill for it. There is no hourly billing and no surprise invoice.

Why this matters to you: the budget you take upstairs is the budget. The risk of the work running long sits with me, not with you.

Milestones

There is no disappearing for six weeks. You see working software as it comes together, and feedback happens while it can still change the build, not at a big reveal that is too late to change.

The dates you are working toward are the ones in the scope document.

A standard milestone and payment schedule is not published on this page. Ask for the one that would apply to your build.

Staging and acceptance

By default, work happens on a staging copy, not on the system your team is using, and you accept it there before it goes live. If a build is set up differently, the scope says so. Until you have said yes, your live system is the one you already had.

Also by default, each release is checked by a second, read-only pass before it is called done. The pass that checks is not the pass that built. Each release is backed up first and leaves a written record of what changed.

One rule sits under all of it. A "nothing found" only counts after the same check has found something real. A check that finds nothing looks exactly like a check that is broken, so the check gets tested before its silence is trusted.

What that looked like, from the field notes

  • A second AI checked a phone app built in one evening. It found 13 problems before anyone used the app, and all 13 were fixed that night. One night by the pool
  • A demo went through three review passes before anyone saw it. One pass found fourteen sentences that said more than the screen could back up. Each was rewritten until it was true. Nine columns and a Monday morning
  • A check passed its first test by finding nothing. So a fake failure was planted to prove the check could see one. The test itself turned out to be wrong, and was fixed. The power went out with a dozen agents running
  • A deleted recording was located with a read-only scan of the drive. The note says plainly what was found and what has not been checked yet. What a fingerprint means

Your data and access

Your data stays where you decide it lives. Where the system runs, and what stays on machines I own versus what uses a hosted AI model, is agreed per project and written into the scope before the work starts. For clients who need it, I deploy on your hardware or your own cloud account. For everyone else, I host it on mine, with a written handover plan.

Here is the instruction from one real restart. When a power cut closed every agent session at once, each new session was told the same three things: read your handoff, check anything it claims is done before you build on it, and send, push, and deploy nothing without a go. The full account.

What access a build needs, and how credentials are held and handed back, is not written up on this page yet. Ask before you share anything.

Handoff and ownership

When the build is done, you own the code, the domain and the data. Where it runs is written in the scope: your account, or mine with a handover plan. No proprietary platform you cannot leave.

In your name
  • The code

    Every line of it, in a repository you hold.

  • Where it runs

    Your account, or mine with a handover plan, written in the scope.

  • The domain

    The address your customers type.

  • The data

    Your records stay yours.

Then it is your call. If you want me to stay and run it, we move to a monthly retainer. If you want to take it in-house, I make sure your team can actually run it.

Why this matters to you: the decision to keep working with me stays yours, with nothing held back to force it.

What happens if I am unavailable

Start with what is already true. You own the code, the domain, and the data, so nothing is held back that only I can open. If it runs on your account, it keeps running without me. If I host it, the scope says how it moves to you, and that handover is the plan to follow if I am ever out of reach.

Work in progress is written down as it happens. Every agent writes its conversation to disk as it goes. When a power cut closed a dozen agent sessions on my workstation at once, nothing was lost, and ten sessions were back nineteen minutes after the machine rebooted, each with a short handoff. Read what came back.

Thirty minutes on September 30

19 minutes

  1. Power cut a little before 2:50 pm
  2. Machine back 2:51 pm, every session closed
  3. First session back 3:03 pm
  4. Tenth session back 3:10 pm

Source: The field note “The power went out with a dozen agents running” on this site.

What this page does not say yet is who would pick up a build in progress if I were out for an extended stretch. That is a fair question for a company your size. Ask it on the first call.

Working with a partner or your existing vendor

Most clients come to me with existing tools, vendors, and workflows. I evaluate what you have, keep what works, and replace what does not. Nothing gets ripped out and rebuilt from scratch unless that is genuinely the right call.

For specialized work, I bring in trusted people I have worked with for years. I still lead every engagement and own every outcome, so you are never re-explaining your business to someone new.

If you run an agency or a software company and want to deliver a build together, the terms for that are not written up here yet. Tell me what you have in mind and what your client needs.

Look at the work before you take my word for it

Systems running today

Named clients, and operational software shown without the client's name, each with how I am related to the people it was built for.

See the work
sample data

A Monday review built on a nine-column log

Change a status and watch every number move. Or paste your own log from Excel. Your log is never uploaded.

Open the live demo

Bring the build you are weighing.

Thirty minutes. Tell me what it has to do and what worries you about it. You leave with an honest read on whether I can help and what it would take.

Prefer email? hello@genesisflowlabs.com