Nidal

@nselmi

What's the Kolmogorov Complexity / Minimum Description Length of a Reasoning Language Model? LLMs, AI/ML, Data Science, Graph DBs. PhD student. 🇹🇳➡️🇺🇸

Lancaster, PA

nselmi.com

Replies Reposts Media

Pinned
Nidal
@nselmi
Apr 15
We will at some point in the future create neural models (much larger than todays) which can run/simulate computers inside them, and then inside those computers we'll be able to run other smaller neural models. Maybe in 3-5 years.
Nidal
@nselmi
Jun 5, 2025
Imagine the 1 Quadrillion model.
1
Nidal
@nselmi
Aug 29
@zoox hear me out
Nidal
@nselmi
Sep 11, 2023
Replying to @nselmi
The driverless future with vintage aesthetics.
1
Nidal
@nselmi
Aug 28
At some point (few years) Postgres servers (maybe even SQLite) might just ship with a highly-efficient Small Language Model. (MCPs were never really a good solution, @postgresml is the closest thing we have so far).
Tinker
@tinkerapi
Aug 27
LLMs with scaffolds have lagged on text-to-SQL, a task that relies on human judgment. By folding expert judgment into every part of RLVR on Tinker, @maxYuxuanZhu and @ddkang (UIUC and Bridgwater) trained the first text-to-SQL model to beat the human mark. thinkingmachines.ai/news/putting-t…
Nidal
@nselmi
Aug 25
A similar thing is happening with cybersecurity where the inference spend for breaking a system is a proxy for its security.
Nidal
@nselmi
Jul 21
Some (unknown) number of mathematical problems are now solvable given a large enough token budget and we're beginning to observe the token (and dollars) cost for finding the truth value of some mathematical statements. In a way, the difficulty of math problems is becoming
Nidal
@nselmi
Aug 24
This is similar to what RL is doing to LLMs and why they seem to write "weird".
Eren Chen
@ErenChenAI
Aug 24
Slow-motion look at the 400m champion’s running form at the World Humanoid Robot Games.
00:00