Umut Ozturk

Helping individuals & organisations
adopt AI in knowledge work

London, UK

What's happening

For most of history, scaling knowledge work meant hiring more people. Then 2017 brought the Transformer. Then GPT-3. Then ChatGPT. By late 2022, it was undeniable: a new kind of cognitive tool had arrived.

What's different this time isn't automation of physical tasks. That's been happening for decades. What's new is AI working on thinking itself: reading, writing, reasoning, deciding. Knowledge work is being fundamentally reshaped. The people and organisations that learn to work with these systems, not around them, will move at a pace that was previously impossible.

How it works (I think)

exp(x) = Σ xⁿ/n! = 1 + x + x²/2! + x³/3! + ...

Small, consistent efforts compound over time. At first, growth feels invisible, but once it hits the steep part of the curve, everything changes. That's the exponential. That's how learning, building, and improving actually work.

AI is on that steep part of the curve right now. The decisions we make today, about how we build and use these systems, compound fast. Small missteps don't stay small for long.

Where I go deep

Most people experience AI as a conversation. But chat is just an interface. Agents are the underlying structures where real work gets done, autonomously and at scale. I'm working on that to make it happen.

01

Large Language Models

The probabilistic core of every agent: how transformers reason over context, where they break down, and what it really means to build on top of them reliably.

02

Context Engineering

Designing what the model sees: prompts, memory, retrieval, tools, skills, MCPs, and plugins so the right information is always in the window.

03

Harness Engineering

The essential infrastructure and systematic discipline required to make AI agents reliable, maintainable, and effective for complex tasks.

What I Work On