Home

NormaEngineering

Go to market

Pre-MVP phase

This phase is how I find out whether anyone wants the answer, before I build it.

I make ad creatives that test the question itself: if I spend this now, can I still cover what comes next. The creatives are there to see which wording people stop for. Ad spend sends the people who stop to a landing page for that question, and the page asks for an email.

What I measure is how many people enter an email. That number is the demand. There is no product tier in this phase. The cost is the ad spend, and I set it when the test runs. The page sits on the site I already run, and the addresses go in a table. A quiet list means the vertical slice doesn’t start.

Vertical slice phase

This phase starts only if those emails say people want the answer. The build is one path through the system, not a finished version of every job. Someone asks “Can I do a $90 dinner Friday?” and Norma parses it, runs the four spend functions, and answers with the reason. Jev classifies the transactions those functions depend on: the recurring bills, and groceries, gas, and other necessities. That classification is part of the slice, because without it the question can’t be answered. It only has to be good enough for this one answer.

The slice is read-only. It can look at the transactions it has and explain a purchase. It can’t move money, and nobody can create their own agent yet. Saved preferences can wait too, because this path looks the numbers up fresh.

The model is expensive, so I still need to learn whether people trust the answer and whether the data stays reliable. Logging in to read transactions, moving money between accounts, new agents, and connecting to banks directly come after this works, and only if people ask for them.