How we work
The inside of the box, week by week.
People buying custom software are afraid of the black box. Money goes in, months pass, something comes out. This page is the box with the lid off. It is the process for a Track A build. Incubate follows the same shape with a partnership agreement in front of it.
- Free, thirty minutes
- One to two weeks
- Three to eight weeks, typical
- Monthly plan
From the first call to month seven.
Week 0
The call
Thirty minutes. You explain what is costing you time. We ask what happens today, who touches it, and what it connects to. By the end we can both say whether it is a fit and roughly what it would take. If it is not a fit, we say so and, where we can, point you somewhere better.
A yes, a no, or a specific question to answer before either.
Week 1 to 2
Scoping
We sit with the people who do the work now. We look at the spreadsheet, the inbox, the workaround. Then we write it up: what it does, what it does not, what it connects to, who can see what, where the data lives, and what we are deliberately leaving out of version one. This is where deciding what to build happens, and it is the part AI did not make faster.
A written spec you can read, and a fixed quote. Both are yours to keep.
Week 3
Foundations first
Before any screen you will recognise, the boring parts go in: accounts and hosting in your name, authentication through a proven provider, tenancy at the database layer, backups with a restore tested, monitoring, logging, secrets on the server, spend limits on every account. This is the week that decides whether the system survives contact with real customers.
An empty app you can log in to, running on infrastructure you own.
Week 4 to 7
The build
This is the part that got fast. Screens, workflows, integrations, in short cycles. You see it running at the end of every week, on a real URL, and you use it. Feedback goes into the next cycle. Scope changes get written down with a price before they happen, not discovered in an invoice.
A working system that the people who will use it have already used.
Week 8
Launch checklist
Real data comes across. We log in as two different users and try to read each other's records. We grep the built code for secrets. We restore a backup. We check the region, the spend alerts, the error tracking, the DNS. We write the runbook. Then we launch, on a quiet day, with someone watching.
A live system, a runbook, and every login in your name.
Month 2 onward
The plan
Patching, monitoring, backup restores, security updates, support, and a bucket of small changes each month. A named person with a stated reply time. The runbook stays current. Larger changes get scoped and quoted the same way the build was.
A system that is still patched, still backed up, and still supported in month seven.
Working rules.
Writing over meetings
Decisions get written down where you can read them later. Meetings are for the things writing cannot do.
You see it running early and often
A real URL from week three. If you have not used it, it is not done.
Scope changes have a price before they happen
Not after. A fixed price only means something if the scope is fixed too.
Everything in your name
Accounts, domain, hosting, code. injuid can be removed in an afternoon.
We say no
To features that do not earn their place in version one, to timelines that skip the foundations, and to work that is not a fit. A no early is cheaper than a yes that goes wrong.
What the AI does, and what it does not.
It does
- Writes the first draft of most of the code, fast
- Generates the tedious parts: forms, validation, data migrations, test scaffolding
- Reads documentation for integrations so we spend less time on it
- Makes it cheap to try two approaches and keep the better one
It does not
- Decide what to build. That comes from scoping, with the people who do the work
- Decide how data is separated, where it lives, or who can see it
- Get to ship anything that has not been read by a person who knows what it does
- Get anywhere near secrets, production access, or your customer data
The tools are good and they are getting better. They are also cheerfully confident about things that are wrong, and they have no idea what a data breach costs. Someone still has to know.