Process
Playbook
Why we ship in weeks, not quarters
Long timelines hide risk. Short ones surface it. Here’s how a six-week engagement stays honest from kickoff to launch.
TAGS
PROCESS
SCOPE
delivery

Marco Vieira
FOUNDER
22 Mar 2026
5 minutes
ON THIS PAGE
01 Fixed scope, fixed price
02 Weekly demos kill surprises
03 The phases stay the same
A six-month build is a six-month bet that you understood the problem correctly on day one. You rarely did. The longer the timeline, the longer the gap between a wrong assumption and the moment you discover it.
We run tight engagements on purpose — usually around six weeks — not because we cut corners, but because short loops keep everyone honest.
Fixed scope, fixed price
Before anything is built, we scope it precisely and price it as a fixed number. No open-ended hourly meter, no surprise invoice. Constraints aren’t the enemy of good work; they’re what force the important decisions to happen early instead of late.
Weekly demos kill surprises
Every week, you see the real thing running — not a status update, not a slide. If we misunderstood something, you catch it in week two when it costs an afternoon, not in week six when it costs the project. There is no big reveal at the end, because by the end you’ve already seen it grow.
The phases stay the same
Discover, build, deploy, support. A week to find the leverage and scope it. A few weeks to build it in the open. A week to ship it properly. Then ongoing support while your team takes ownership. The rhythm is the same whether it’s an audit or an end-to-end system — only the middle stretches.
Speed isn’t the goal. Honesty is. Short timelines just make it harder to hide.
archivio { 1 }
OTHER POSTS
Read related content
Read our posts on strategy, engineering, process, opinion and field notes.
RELATED INSIGHTS
All
Strategy
Engineering
Process
Opinion
Field notes

Context windows aren’t memory
Stuffing everything into the prompt feels like memory, but it isn’t. Real systems decide what to retrieve, when, and what to forget — here’s how we think about it.
Inês Carvalho
6 minutes
Read post

Most "AI problems" aren’t AI problems
Before you build an agent, it’s worth asking whether the bottleneck is the model — or the broken process around it. Usually it’s the process.
Sofia Almada
5 minutes
Read post

What "production-grade" actually means for an AI agent
A demo proves an agent can work once. Production means it works on the thousandth strange input, at 3am, when nobody’s watching. Here’s the gap between the two.
Inês Carvalho
7 minutes
Read post

The handoff is the product
A system you can’t run without us isn’t an asset — it’s a dependency. We’d rather build something you own outright.
Marco Vieira
4 minutes
Read post

Agents vs. wrappers: a field guide
“We added AI” can mean a thin wrapper around a chat box, or a system that does real work end to end. The difference matters more than the label.
Inês Carvalho
6 minutes
Read post
