necromancer

very gentle

[DHH @dhh]

These stochastic parrots are getting awfully good at doing things nobody taught them to do. Generalization!

[Mistral AI @MistralAI]

Meet Mistral Large 4, aka Le Chonk.

• 1T parameters, natively multimodal. 49B active.
It is the best open weights model from US or Europe on aggregated benchmarks.
• State-of-the-art on critical workloads, including cyber defense, manufacturing and finance and it surpasses closed frontier models on visual grounding.
• Forged in Europe end-to-end and is deployable from Europe via our own Mistral Cloud infrastructure.
• Available to all via API today. Working with cybersecurity partners privately.

Open weights release end of October.

[necromancer @aflatooun]

@dhh assembly is faster

[necromancer @aflatooun]

@MonicaLMarks find it hard to believe

[necromancer @aflatooun]

@miniapeur ya. i tried to read it but couldnt bcz it got so komplex.

[Yann LeCun @ylecun]

@PessimistsArc Right. Dario was already claiming that GPT2 was too dangerous to open source back in 2019.
I made fun of them then.
Everyone should make fun of them now.

[necromancer @aflatooun]

i find llms to be a goldmine for creative ppl. you can literally do anything you want. instead of fearing AI, we shud embrace it.

[necromancer @aflatooun]

@kirancodes they are coming for everyone

[doomslide @doomslide]

Academics live perilous lives already. They can't afford but to flee when the machine approaches their domain. Unless we see serious regulation one way or another this is likely the last decade of open science in the west. 3/3

[Sheikh Hopeful @high_hopeful]

From helping Pakistan exit the FATF grey list to being sent on a 20 days remand

[necromancer @aflatooun]

@alexandr_wang need it in cursor

[necromancer @aflatooun]

@Eng_Hemdi all fields are dead. only robotics and doctors survive for now.

[brief. @brief_pk]

Islamabad has built a self-dealing procurement loop: state and military-linked entities capture no-bid contracts, pocket the margins, and cheaply subcontract the work back to shut-out private firms. Backed by SIFC’s Article 10F and exposed under Rule 32-F, the mechanism is now under pressure as the IMF demands a binding 40% cap on subcontracting.
https://t.co/oVGpGx4NOL

[Danyal Rasool @Danny61000]

Pakistan announce a huge cull, with 7 players recalled from England. Rizwan, Agha, Imam, Ali Usman, Khurram are among them, while Saim Ayub and Abdullah Fazal are among the replacements.

Coaches Sarfaraz and Umar Gul have also been recalled, while Hesson and Noffke take over.

[John Arnold @johnarnold]

One day it'll come out that the CIA infiltrated Western consulting firms and convinced Saudi to blow its wealth on LIV Golf, The Line, domestic soccer, Lucid, a ski resort, an artificial lake, World Cup stadiums & a giant cube instead of converting it into geopolitical influence.

[necromancer @aflatooun]

deepseek v4 pro is very good at writing

[Rana Ayyub @RanaAyyub]

On Imran Khan. The man Pakistan can jail but not erase

Written earlier this year

https://t.co/F5YMqo9I5Y

[Nathan Lambert @natolambert]

My book, Reinforcement Learning from Human Feedback is done!

This is the book I wish I had when learning to fine-tune, align, & now post-train models since ChatGPT. The resource has been built by me finding time to study and document the fundamentals on nights and weekends since 2024.

Transferring as much of the intuitions of building Olmo as I possibly can in the book format.

The book is launching with an over 10 hour, full course with slidedecks, functional code for the training chapters, an example model completions library, and of course the free online web version.

Physical orders from Manning will ship in 1-2 weeks, and Amazon a week or so after. Thanks for your support!

[Ejaaz @cryptopunk7213]

kimi k3 feels more significant than the deepseek moment. hard to unsee it:

> beats fable, got 5.6 on multiple key benchmarks. 1st time a chinese model *doesn’t* trail

> it does so for a fraction of the size, cost and without the swanky nvidia gpus

> repeated breakthroughs in data used to train models + training techniques have made china a legitimate centre for ai research

all of this is open source too. at what point do we start acknowledging they might actually have some advantage here?

gap has closed to 3 months tops now.

[Javier Blas @JavierBlas]

Marco Rubio (June 23): "No country is allowed to charge tolls or fees on an international waterway"

Donald Trump (Jul 13): "The U.S.A. will be [...] known as 'THE GUARDIAN OF THE HORMUZ STRAIT,' but as such [...] will be reimbursed, at the rate of 20% on all cargo shipped"