AI at work. What worked, what didn't. Written by Martynas Kairys.

Practical notes on artificial intelligence in an ordinary office job.

What you must never paste into ChatGPT: the one-page rule for employees

New here? Where to start.

Where to start

Latest

all notes

From the lab

all experiments →

Frameworks

  • AI as an x-ray AI as an x-ray AI does not break anything. It just shows where the work was never there in the first place.
  • Automating tasks or augmenting decisions Automating tasks or augmenting decisions Two completely different things go by the same name. Which one you pick decides whether experiments ever become an operating model.
  • Rangers, not narrow specialists Rangers, not narrow specialists Ten thousand hours in one narrow field is now a risk, not a shield. What to grow instead.
  • Vibe coding with limits Vibe coding with limits You can build software in plain words without an engineering degree. But only inside a drawn boundary, and only if you are stubborn.

all entries

In pictures

  • The Monday after the holiday The Monday after the holiday A comic about the first morning back from leave: forty three unread emails, an AI summary in two minutes, and an agent that swapped two columns.
  • The contract that left the building The contract that left the building A comic about one Monday morning: forty pages, two keys, and a document that left the company before its owner did.

all entries