pillka.hn

AI · Consulting · Labs

I advise on AI, and I run it myself.

This is what comes out of it: a robot rover, a cluster of five AI agents, an agent that makes its own investment decisions, and a simulation that works through policy choices ten years ahead. Not case studies on slides, but running systems, limits included.

AI-generated image (Higgsfield)

Lab notebook

Each lab is an experiment with an open outcome. The stamp shows where it stands.

  1. Rover Lab

    A robot rover as a platform for AI where it meets the physical world.

    in progress
    AI-generated image (Higgsfield)
  2. Beatles Lab

    Five AI agents, one orchestrator, daily work rather than a demo.

    running
  3. Paul the Investor

    An autonomous agent makes investment decisions and learns from them.

    running
  4. nowplus10

    Germany in ten years, as a consequence of today's decisions.

    in progress
  5. Flipped-Foresight (PicturesOfTomorrow) ↗ (opens external site)

    Scenarios and pictures of the future with AI, as an application in use.

    running

What I have learned

  1. Agents need an architecture, not a chat.

    Bot-to-bot messaging over chat apps does not work by design. The agents in the cluster are addressed directly, each with a clear role and one orchestrator.

  2. Security starts with separation.

    Agents run in their own isolated user accounts and reach vehicles and machines only through a private network. None of it can be reached from the internet.

  3. The AI tells the story, the model does the maths.

    Where numbers have to hold, they belong in an open, testable model. Language models explain, research and write, but they do not estimate consequences.

Consulting

The labs have turned into an offer for organisations that are serious about using AI.

Formats and approach →

Contact

Email is the easiest way. I reply personally.

prof@pillkahn.com