Skip to main content

How we work

How we build, in numbers you can check

A snapshot taken from our own Git and GitHub history. We publish what holds up and say plainly what we do not measure yet.

Snapshot as of 25 October 2026

  • 5,200

    commits across 7 repositories (4 shown below) since November 2025

  • 3,400+

    pull requests merged, with automated dependency updates subtracted

  • 1,300+

    commits in the last 30 days

  • 288

    automated test functions in Parley, our open-source library

Four projects, one measured the same way

Project Since Commits Merged PRs Passing CI runs on main
futurist.systems 2025-11 940 786 92/96
Parley 2026-07 927 117 95/99
Expato 2026-07 1,046 1,000+ 95/95
Valivans Club 2026-02 670 466 20/20

Successful CI runs are counted among the last 100 or so runs on the main branch. A count marked with a plus hit the 1,000-result limit of our query, so the real number is higher.

How a change reaches production

We build with a fleet of AI coding agents working in parallel. By policy, changes to this site go through the same four steps, whoever or whatever wrote them:

  1. 01

    Local multi-angle review

    Before a pull request is opened, separate reviewers check the diff for security, performance, tests, clarity and edge cases. Blocking findings are fixed first.

  2. 02

    Automated checks on every push

    Lint, type check and a full build run before code leaves a developer machine, and again in CI.

  3. 03

    A second review by someone other than the author

    A reviewer other than the author must clear the exact commit that ships: normally an automated code reviewer, or a documented local multi-angle review when it is unavailable. A new push needs a new review.

  4. 04

    A merge check, never a direct push

    Nothing is pushed to the main branch. A script merges only when review and CI pass, plus a preview check for visual changes.

What you can inspect yourself

What we do not claim

  • No uptime figure. We have no long monitoring history yet, so we do not quote one.
  • No lines-of-code count. Agents write code fast, so volume says little about quality.
  • Not every client project is listed. Some repositories are private, and we only name work whose owners agreed.
  • Commit and PR counts show activity, not value. The checks above are what turn activity into trust.

Questions about these numbers

Where do the numbers come from?

From the Git history and GitHub pull request and CI records of our own repositories, read on 25 October 2026. They are a snapshot, not a live counter.

Do AI agents write your code?

Yes, a large part of it, working in parallel. People set the goals, review the result and own every decision. The review steps above apply to agent-written changes exactly as it would to anyone else's.

Can I see the code?

Parley and our Claude Code rules repository are public. Client and internal repositories are private, and we do not name or show them without the owner's agreement.

Why is there no uptime number?

Because we do not have enough monitoring history to make an honest claim. We prefer to leave it out than to estimate it.

Ready to put AI to work?

30-minute call. No commitment. Let's explore what AI agents can build for you.

Free 30-min consultation
Custom project roadmap
No obligation
Based in Valencia, Spain • EU timezone • English, Spanish, Ukrainian, Russian
Languages built-in