Sofus Mortensen

[Sofus Mortensen @sofusmort]

@ChShersh “Hello, World”

[Sofus Mortensen @sofusmort]

@ChShersh Thanks to AI there’s still zero reason to learn Python.

[Sofus Mortensen @sofusmort]

@HowToAI_ Deterministic breakthrough was 1999 with SSSP in linear time.

https://t.co/R89cPdCXmR

[How To Prompt @HowToPrompt__]

Yann LeCun was right the entire time. And generative AI might be a dead end.

For the last three years, the entire industry has been obsessed with building bigger LLMs. Trillions of parameters. Billions in compute.

The theory was simple: if you make the model big enough, it will eventually understand how the world works.

Yann LeCun said that was stupid.

He argued that generative AI is fundamentally inefficient.

When an AI predicts the next word, or generates the next pixel, it wastes massive amounts of compute on surface-level details.

It memorizes patterns instead of learning the actual physics of reality.

He proposed a different path: JEPA (Joint-Embedding Predictive Architecture).

Instead of forcing the AI to paint the world pixel by pixel, JEPA forces it to predict abstract concepts. It predicts what happens next in a compressed "thought space."

But for years, JEPA had a fatal flaw.

It suffered from "representation collapse."

Because the AI was allowed to simplify reality, it would cheat. It would simplify everything so much that a dog, a car, and a human all looked identical.

It learned nothing.

To fix it, engineers had to use insanely complex hacks, frozen encoders, and massive compute overheads.

Until today.

Researchers just dropped a paper called "LeWorldModel" (LeWM).

They completely solved the collapse problem.

They replaced the complex engineering hacks with a single, elegant mathematical regularizer.

It forces the AI's internal "thoughts" into a perfect Gaussian distribution.

The AI can no longer cheat. It is forced to understand the physical structure of reality to make its predictions.

The results completely rewrite the economics of AI.

LeWM didn't need a massive, centralized supercomputer.

It has just 15 million parameters.

It trains on a single, standard GPU in a few hours.

Yet it plans 48x faster than massive foundation world models. It intrinsically understands physics. It instantly detects impossible events.

We spent billions trying to force massive server farms to memorize the internet.

Now, a tiny model running locally on a single graphics card is actually learning how the real world works.

[Sofus Mortensen @sofusmort]

@muellerberndt Behold your maker.

[Sofus Mortensen @sofusmort]

@ChShersh No. On the contrary it’s a great sign of confidence and maturity that you ask questions about what you simply just don’t understand - instead of asking a wiseguy question purposed to make you look really smart.

[Sofus Mortensen @sofusmort]

@thorborg Schweiz har formueskat - men til gengæld er skat på kapitalindkomst oftest 0.

Det gør Schweiz meget attraktivt for iværksættere.

Min pointe er at det ikke formueskat i sig selv der er et problem. Tværtimod. Problemet er den øvrige beskatning.

[Sofus Mortensen @sofusmort]

@thorborg Schweiz har formueskat.

[Sofus Mortensen @sofusmort]

@ChrHallum Ja tak. Kopier skattesystem fra Schweiz

[Sofus Mortensen @sofusmort]

@BoWang87 “Nuclear launch detected”

[Sofus Mortensen @sofusmort]

@JustinWolfers Before any of you get really upset and hurt about this, please understand that Justin is not laughing with you.

[Sofus Mortensen @sofusmort]

@karabaye14 "O Captain! My Captain!"

[Sofus Mortensen @sofusmort]

@ChShersh Of course FP doesn’t magically fix bugs - that’s the promised land of dependent types. lol.

But what about how easy it is for others to read the code and assess if it is working or not?

[Sofus Mortensen @sofusmort]

@ChShersh *to think you finally understand

[Sofus Mortensen @sofusmort]

@ChShersh Great read

[Sofus Mortensen @sofusmort]

@rasmalai Fortran
Haskell
Ada
Prolog
LISP
F#
ALGOL

[Sofus Mortensen @sofusmort]

@RpsAgainstTrump Mac’n cheese guy.

[Sofus Mortensen @sofusmort]

@davepl1968 @iluminatibot Hey @grok. Where are you more likely to die from atherosclerosis - considering quality and availability of healthcare - Canada or USA?

[Rasmus Jarlov @RasmusJarlov]

Every insult, threat, tariff and lie that we receive strengthens our resolve.

The answer from Denmark and Greenland is final:

We will never hand over Greenland.

We pray that our true allies will stand with us because we are going to need it.

[Sofus Mortensen @sofusmort]

@ChShersh @JayD0ubleu Me too. You’re a better engineer understanding what your tools produce and how they do it. Applies when using a C++ compiler and even more so when relying on AI to do your work.