For Startups · 5 pages

AI for startups.
Shipped in weeks.

Startups do not have a procurement process, they have a burn rate and a board meeting. We scope work to fit both: a working system in weeks, priced against the milestone it unlocks.

Shipped & in build · 13

Our own · coming soon

yourOS

An AI operating system, with Roxy at its heart.

Live · iOS · Android

Sonora

AI sound therapy for Cognitive AI.

Five platforms

Podswap

A growth engine with no human in the loop.

Prototype · closed beta

unOpen.AI

A private assistant with permanent memory.

Live

Coherence

Identity and motion, generated with agents.

Live

Domain Booth

A domain brokerage, rebuilt from the mark up.

Domain landing page · Built

Eclipse.com

One day, one lunation

Domain landing page · Live

ZeroOne.com

The whole page is a switch

Domain landing page · Live

Altitude.com

A climb through the atmosphere

Domain landing page · Built

Coral.com

Field notes on a living word

Domain landing page · Live

Ascend.com

Registered in 1990, once Nokia’s

Domain landing page · Built

Protocol.com

The route is the P

Domain landing page · Built

Engineer.ai

The one-word .ai for coding

01 / 13

All our work →

At a glance

Scoped to the milestone.

01 What you get: A working system on real data within weeks, or a technical proof built to survive diligence rather than a stage demo.
02 How it is priced: As fixed sprints against a written outcome, so the number on the invoice is the number agreed at the start of the sprint.
03 Who owns it: You do. Code ships into your repositories and cloud accounts from day one, with handover documentation written as part of the work.
04 When we say no: If an off-the-shelf tool and a week of integration settle the question, we will say so on the first call.

WHAT STARTUPS ACTUALLY NEED

Three things before the first hire.

01

Something real to show

A demo that survives a hostile question from a technical investor. Not a Figma flow, not a wrapper around a single prompt — a system that does the thing on real data.

02

Senior judgement, briefly

The decisions that matter in an AI product get made in the first three weeks: retrieval strategy, evaluation approach, where the model boundary sits. Getting them wrong is expensive and hard to see until later.
03

To not hire yet

A senior AI engineer is a long search and a large permanent cost. Before product-market fit, borrowing that capacity is usually the better trade.

HOW WE WORK WITH STARTUPS

Clear scope, honest estimates, code you keep.

Time-and-materials suits the agency and nobody else when runway is the binding constraint. We scope startup work as fixed sprints against a written outcome, so the number on the invoice is the number agreed at the start of the sprint.

Everything ships into your repositories and your cloud accounts from day one. There is no proprietary platform, no runtime dependency on us, and no licensing that turns into a line item forever. If your first engineering hire wants to take it over in three months, that is a supported outcome — we write handover documentation as part of the work rather than as a paid extra.

We will also tell you when a project does not need us. A meaningful share of “we need an AI feature” conversations resolve to a well-chosen off-the-shelf tool plus a week of integration, and saying so on the first call costs us a project and earns the next three.

PICK THE MILESTONE

Start from the milestone, not the tool.

Each milestone is scoped as fixed sprints against a written outcome. Pick the one your next board meeting, pilot or round depends on.

Four milestones

Investor proof · ~3 weeks

A board demo that survives questions

A demo that survives a hostile question from a technical investor. Not a Figma flow, not a wrapper around a single prompt — a system that does the thing on real data.

Timing: ~3 weeks
Runs on: Real data, not a single prompt
Not: A Figma flow or a stage demo

1Built to survive a technical investor’s questions
2Works on real data, not a wrapper around one prompt
3Code ships into your repositories and cloud accounts from day one

Investor proofs

AI MVP · ~6 weeks

An MVP for the first pilot customer

A working product slice, on real data, that real users can try. Scoped to test one hypothesis.

Timing: ~6 weeks
Scope: One hypothesis, tested with real users
Team: Pairs with your own engineers

1Scoped to test one hypothesis
2A working product slice on real data that real users can try
3The AI-specific decisions are transferred to your team deliberately: retrieval, evaluation, model selection, cost control

AI MVP development

Investor proof · ~3 weeks

Investor proof for the round

A demonstrable technical proof built to survive diligence, not just a stage demo.

Timing: ~3 weeks
Built to: Survive technical diligence

1A demonstrable technical proof, not a stage demo
2Code ships into your repositories as it is built
3Sprints are short, so a change of direction costs one sprint rather than a renegotiation

Investor proofs

Monthly · Fractional team

Senior capacity without a permanent hire

Ongoing senior capacity alongside your own engineers, without a permanent hire.

Engagement: Monthly
Works: Alongside your own engineers
Aim: Self-sufficient within a couple of months

1Before product-market fit, borrowing senior AI capacity is usually the better trade than a permanent hire
2Retrieval strategy, evaluation approach and the model boundary are decided early, then transferred deliberately
3Teams that pair with us are meant to be self-sufficient within a couple of months, and we treat that as the successful outcome

Fractional AI team

FOR STARTUPS — ALL PAGES

Every startup page.

STU.01

AI MVP Development

A working AI MVP in around six weeks, scoped to test one question with real users.
Open page
STU.02

Fractional AI Team

Senior AI engineering capacity without a permanent hire, for teams between stages.
Open page
STU.03

Prototype to Production

Taking a demo that works on stage and making it survive real traffic and real data.
Open page
STU.04

AI for Non-Technical Founders

What a non-technical founder needs to understand to make good AI build decisions.
Open page

Building software since: 2006
Based in: Barcelona
Model stance: Agnostic
Code ownership: Yours

Facts about how we work, not measured client outcomes.

Questions

What founders ask before the first call.

What does an AI MVP actually cost?

A focused six-week MVP typically lands in the €30–60k range depending on integration count and how much data preparation is needed. An investor or board demo is smaller — usually €8–20k over two to three weeks. Our pricing page carries the current bands — rough estimates from past work, not fixed prices. If your budget is materially below that, say so early: sometimes the scope can be cut to fit, and sometimes the honest answer is that we are the wrong shape of vendor for you right now.

Do we own the code?

Entirely. It ships into your GitHub organisation and your cloud accounts as we build, not in a handover at the end. Infrastructure is defined as code so it is reproducible without us. There is no platform fee, no licence and no runtime dependency on our infrastructure.

Can you work alongside our own engineers?

That is usually the best arrangement. Your team knows the product and the domain; we bring the AI-specific decisions — retrieval design, evaluation, model selection, cost control — and transfer them deliberately. The aim is for teams that pair with us to be self-sufficient within a couple of months, and we treat that as the successful outcome.

What if we pivot mid-engagement?

Startups pivot; that is not a failure mode to be penalised. Sprints are scoped in short increments precisely so a change of direction costs one sprint rather than a contract renegotiation. We would rather re-scope at a sprint boundary than deliver something you no longer need.

Are you going to build us a thin wrapper around an API?

If that is genuinely all your problem needs, we will tell you and you should not pay agency rates for it. What justifies engineering is everything around the model call: retrieval over your own data, evaluation, cost control at volume, handling the cases where the model is wrong, and the product surface that makes it usable. That is where the difficulty and the defensibility live.

Hire your first
AI employee.

Tell us the job you would hand over if you trusted a machine with it. In one 30-minute call we will tell you whether it should, what it would take — and if it should not, we will say so.

Pick the job — the form arrives ready for it

Book a 30-minute call How we work

30-minute technical call · no deck

Services

Our Work

Agents We Run

For Enterprise

For Startups

By Industry

Technology

Hosting & Infrastructure

Trust & Security

Comparisons

Reference Architectures

Resources

Where We Work

Company

AI for startups.Shipped in weeks.

Scoped to the milestone.

Three things before the first hire.

Something real to show

Senior judgement, briefly

To not hire yet

Clear scope, honest estimates, code you keep.

Start from the milestone, not the tool.

A board demo that survives questions

An MVP for the first pilot customer

Investor proof for the round

Senior capacity without a permanent hire

Every startup page.

AI MVP Development

Fractional AI Team

Prototype to Production

AI for Non-Technical Founders

What founders ask before the first call.

Hire your first AI employee. AI employee. night shift. quietest teammate. tireless analyst. unfair advantage.

Real work. Check it yourself.

yourOS

Sonora

Podswap

unOpen.AI

Coherence

Domain Booth

Twenty-one years. Names you know.

AI for startups.
Shipped in weeks.

Hire your first
AI employee.