01 / 06 · ORIGIN

Emir Alanyalioglu

I run my companies on AI agents.

I build AI products, run my teams on AI agents and show the systems. One at a time, in the Lab.

Role
Head of Technology · Inteliate
Venture
Co-founder · Fabrikode
Since
Building AI systems since 2020

A replay of a published entry. Not a live feed.

02 / 06 · WHERE IT RUNS

Where it runs

The proof runs every day.

Two companies. One I lead engineering for, one I co-founded. In both, the engineering runs on agent systems I built and operate.

01

Inteliate

Head of Technology

I lead the engineering team. We build AI platforms for security-sensitive environments. The team runs on agent systems for task routing, review and release gating. No clients, no screenshots, on purpose.

  • task routing
  • code review
  • release gating
02

Fabrikode

Co-founder · Engineering software

Engineering software for manufacturers. Three products, two people, AI-assisted development end to end. The company page writes itself from the commit history.

  • WaterForm
  • DragAero
  • ThermoForm

03 / 06 · FILTER

Eliminate the noise.
Engineer the system.

I have been building AI systems since 2020, long before the ChatGPT wave. The rules below come from running them every day.

04 / 06 · RULES

How I work

Four rules I run on.

  1. I

    It runs in my companies first.

    Nothing goes into the Lab until it has run in my own work. Including the week it broke.

  2. II

    The agent writes. I decide.

    Agents draft, record, review and cut. A person reads everything before it ships.

    LAB 01
  3. III

    Not everything deserves a pipeline.

    I automate the step that repeats and changes. I click through the step that does not.

    LAB 02
  4. IV

    Own what you cannot lose.

    If a vendor disappears, the system should still run. The parts that matter, I own and document.

    LAB 03

05 / 06 · OUTPUT

The Lab

One system a week.

Every entry is one system: the problem, how I solved it and what broke. New entries come with a short video, and the template when there is one.

  1. 02

    Not everything deserves a pipeline.

    I automated the recording, the scripts and the editing of eight training videos. The voice-over I generated by hand: eight clicks. Wiring the API would have taken longer than the clicks.

    · Engineering teamWritten

06 / 06 · TERMINAL

Follow the work

Watch it get built.

New entries land here first. Short clips go to LinkedIn, X, Instagram and YouTube.