Skip to content
ARANEUX

How we work

How a project runs here

Four steps, every time. We keep the loops short, get real users in early, and the same person leads from the first call to the handover.

  1. 01 · Week 1

    Watch the work

    Before any code, we sit with the people who do the job today. Who touches it, where it breaks, what done looks like. For AI work we collect twenty real examples, and those become the test set.

    • A map of the workflow
    • Success measured in numbers
    • Real examples to test against
  2. 02 · Weeks 1 to 2

    Write a tight scope

    A short written plan: what we'll build, what we're leaving out, how we'll know it works and what it'll cost to run each month. Vague scope sinks more AI projects than bad models do.

    • Written scope
    • Fixed estimate
    • Architecture and model choice
    • Monthly running cost
  3. 03 · Weeks 2 to 10

    Ship in thin slices

    Small releases in front of real users, early and often. You get a working demo every week. AI features run next to your team with a person approving each step until the logs say it's safe to let go.

    • Weekly releases
    • Eval suite
    • Logs and dashboards
    • Code in your repo
  4. 04 · Ongoing

    Stay for the messy part

    Live systems break in ways demos never do. We handle the debugging, the model upgrades, the integration nobody documented. Then a clean handover to your team, or a monthly retainer.

    • Monitoring and on-call
    • Handover docs
    • Team walkthrough
    • Optional retainer

Principles

What stays the same on every project

  • One technical lead

    The person who scoped it builds it. You'll never explain your business twice.

  • Evals before anything ships

    Every AI change re-runs against your real cases, so a fix in one place can't quietly break another.

  • A person approves the risky stuff

    Anything expensive or customer-facing waits for a human until the data shows it doesn't need to.

  • Boring stack, on purpose

    Next.js, Python, Node, Laravel and Postgres. Your next engineer already knows them.

  • Cost and speed budgets

    We track token spend and response time per feature, so the monthly bill never surprises anyone.

  • Short updates, no chasing

    Trello or Jira, a weekly demo, WhatsApp for quick questions. You always know what's done and what's next.

Start here

Got a workflow that eats your team's week?

Tell us about it. Hussain reads every message himself and replies within a working day, usually with a question or two.