Daniel Paleka’s Newsletter
Subscribe
Sign in
Home
Archive
About
Latest
Top
Discussions
The math that OpenAI keeps for itself
OpenAI has recently released more than 700 mathematics results. What sort of math they didn't publish?
Oct 16
•
Daniel Paleka
55
6
3
December 2025
Writing in public is still underrated
If you have ideas but never write them down, this post is for you.
Dec 1, 2025
•
Daniel Paleka
51
3
2
November 2025
Underrated reasons to be thankful: AI research edition
I celebrated my first Thanksgiving and had to write this.
Nov 28, 2025
•
Daniel Paleka
18
Even superhuman AI forecasters are only as good as your questions
Predicting the future is difficult. It might become easier. What then?
Nov 27, 2025
•
Daniel Paleka
10
1
2
What’s the deal with RL and forecasting?
Prediction is difficult, especially about the future.
Nov 20, 2025
•
Daniel Paleka
6
1
Contra Dwarkesh on RL sample-efficiency via information theory
Supervised learning teaches the model more bits/sample than RL; but it's not the right way to think about it.
Nov 18, 2025
•
Daniel Paleka
50
4
2
16 questions about reinforcement learning
AlphaProof constructed a large curriculum of tasks by randomly misformalizing math olympiad problems; creating many problems of varying difficulty, from…
Nov 18, 2025
•
Daniel Paleka
6
3
Obvious ways RL can fail
Reinforcement learning works for some things and not for others. Why?
Nov 15, 2025
•
Daniel Paleka
3
The two types of LLM preferences
The standard approach to measure values or preferences of LLMs is to:
Nov 10, 2025
•
Daniel Paleka
18
1
October 2025 AI safety news: Adaptive attacks, Tokenization, Impossible tasks
These days, I imagine it is rough for researchers working on LLM defenses.
Nov 7, 2025
•
Daniel Paleka
3
You are going to get priced out of the best AI coding tools
The best AI tools will become far more expensive
Nov 5, 2025
•
Daniel Paleka
29
8
2
A/B testing could lead LLMs to retain users instead of helping them
OpenAI’s updates of GPT-4o in April 2025 famously induced absurd levels of sycophancy: the model would agree with everything users would say, no matter…
Nov 2, 2025
•
Daniel Paleka
10
8
This site requires JavaScript to run correctly. Please
turn on JavaScript
or unblock scripts