~Far be it from me
or, views from my ascent of the shoulders of a giant tower of people
Writing is thinking. There is a lot to think about, therefore there is a lot to write about. Here is a collection of writings reflecting snapshots of my progress in thinking.
While I aim for readability, I don’t aim to be read. Beware that your time might be better spent elsewhere. If you do spend time here, I welcome comments/questions via email.
“Planned” essays are not promised.
§Essays on technology and civilisation, especially AI risk
For thousands of years men dreamed of pacts with demons. Only now are such things possible.—Michèle of the Turing Police, Neuromancer
Cracks in the foundation of the world:
- Move fast, break everything (planned)
- Actually aiming at alignment (planned)
- Simplicity and convergent instrumentality (planned)
Research agendas:
- What is the effective objective? (planned)
- Model enumeration (planned)
- The Honmoon agenda (planned)
Guest lecture:
- Ethics and the future of intelligence (2024.05.23)
Open letters:
- Contra Zuckerberg on ‘Open Source AI’ (2024.07.26)
- Safe and responsible AI in Australia (2023.07.28)
Fiction:
- One size fits all (2026.01.02)
§Notes on the theory of computation and learning
Goal misgeneralisation:
- Instrumental/intrinsic value ambiguity (2025.07.25)
Learning singular learning theory:
- The tempered posterior via the principle of maximum entropy (2026.02.27)
- Blowing up (2025.08.10)
- Schwartz distributions (2025.08.17)
Cheating Kolmogorov sequence:
- Turing trees (2025.06.15)
- Kolmogorov complexity via Turing trees (planned)
- What is complexity? (2026.03.16)
- Embedded computation (planned)
- Hacking the teacher’s password (planned)
Transformers:
- Transformers are universal in-context function approximators (2024.12.05)
- Attention: All you need to know (2025.03.29)
- Transformers: Perceptrons in disguise (2026.03.05)
Decision theory:
- Minimax, three ways (2025.02.23)
- Approximate minimax, three ways (2025.02.25)
Expectiles are underrated sequence:
- Expectiles are configurably-optimistic expectations (2025.06.11)
- Expectiles are simple to compute (2025.06.11)
§What passes for my research blog
Research journal:
- Seven saturdays with singular learning theory (2023.07–unfinished)
- Volumetric analysis of ICLR 2025 review process (2024.12.08)
- Freeing 118 GB of W&B experiment data from a broken binary format (2025.12.29)
- Switching to JAX (2026.03.01)
Goals for my graduate studies:
- Balanced academic orbit (2025.05.25)
- Drinking from the firehose (planned)
Quotation collections:
- On how to choose what to work on (2026.03.16)
§About the author
Rules I try to live by:
- Smile! (2025.07.15)
- Fear of missing out (planned)
What I have been reading (and select detailed book summaries):
- Generalised readings, 2019–2025 (2026.02.01)
- Generalised readings, 2026
(2026.01.31–)
- Replacing Guilt by Nate Soares (2026.03.07)