Oly on AI

A (Slightly) Mechanistic Theory for Exponentially Increasing AI Time Horizons?

AI ‘time horizons’ are mostly not about time (I think it’s mostly ‘data’, but you’ll see where I’m unsure)

May 24 • Oliver Sourbut

6

3

Engineering a Safer World

Risk Modelling — and Safety Engineering? — for AI Loss of Control

May 17 • Oliver Sourbut

7

6

1

Reinforcement learning scaling might incentivise hidden reasoning architectures for AI

Another reason really competent AI might lie to you

May 10 • Oliver Sourbut

2

2

How did ‘large’ language models get that way?

The role of Transformers and Pretraining in GPT

May 3 • Oliver Sourbut

2

1

April 2026

“Best humans still outperform”

One turning point in the history of cope around artificial intelligence

Apr 17 • Oliver Sourbut

4

1

Orders of magnitude: use semitones, not decibels

How I use a secret part of my brain for forbidden mathematics

Apr 1 • Oliver Sourbut

1

1

February 2026

January 2026

The First Type of Transformative AI?

Ask 'what will AI transform?', not 'when?'... and what choices do we have in that?

Jan 6 • Oliver Sourbut

7

December 2025

A Full Epistemic Stack

Knowledge Commons for the 21st Century

Dec 19, 2025 • Oliver Sourbut and Ben Goldhaber

9

4

July 2025

May 2025

You Can’t Skip Exploration

Why understanding experimentation and taste is key to understanding AI

May 21, 2025 • Oliver Sourbut

2

4

April 2025

Is the Cat Out of the Bag?

Who knows how to make AGI?

Apr 10, 2025 • Oliver Sourbut

1