Hamza Mahmood

I play around with weights and activations of LLMs

[alexei @alexeixbt]

it might sound woo woo, but radiating positive energy is the closest thing to magic i've found. tip generously, compliment strangers, share your knowledge with no expectation of anything in return. give like you have more than enough, and eventually you’ll find that you do.

[Hamza Mahmood @_hmahmood]

I did something similar last year with llama 8B model, changed a neuron to update/delete targeted factual knowledge. I should prolly write something about it

[lyra @_lyraaaa_]

if you have a qwen3.5 4b around, please try steering neurons L14 N8949 +1 and L13 N7956 -1
now you have evil qwen

recipe: dose × median(‖resid‖ leaving layer L, tokens 1:) / ‖W_down[:, n]‖` at the down_proj at every input

[Bryan Cheong @bryancsk]

You're telling me Jewish folks thought if you arranged words and symbols systematically in the right way you can create intelligent golem entities and the two companies they founded to do this are worth several trillion?

[Vals AI @ValsAI]

For decades, researchers have sought materials that sort electrons by spin while their magnetism cancels.

In 3 days, 90+ Opus 5.5 agents helped us uncover two room-temperature magnetic semiconductor candidates in simulations: YBaMnFeO₅ and KV[Cr(CN)₆].

KV[Cr(CN)₆] was synthesized back in 1999. Its predicted ability to sort electrons by spin appears to have been hiding in plain sight for 27 years.

[Robin Bilgil @RBilgil]

Made a real-time slop detector with jev as you scroll

[Emad @EMostaque]

This is basically an Opus 4.5/4.6 level model that will work on anything with 8 Gb RAM

[Diogo Almeida @CompleteSkeptic]

After co-inventing ChatGPT, I kept asking myself: why have superhuman chat models not led to AGI?

I’ve spent the last 2 years in stealth building a new way to train models (RLCD), and a new type of frontier AI model that we are releasing today: Jev

• 20-200x faster
• 40-400x cheaper (w/ output tokens free)
• Frontier composable intelligence optimized for decisions

AFAICT the shortest path to AI-based economic revolution

[Shannon Sands @max_paperclips]

what on earth is this?

[Jason @jsonnelson]

“MIT wants to rebuild education around the things AI can't replace. That means oral exams, semester portfolios, in-person project work, and a required social component in every subject. The report even floats the idea of rethinking grades entirely, since without a GPA to optimize, much of the incentive to cheat with AI evaporates.”

[Tony Montana @9mmScorpion]

Countries safe from israel’s Jericho IIII

[Hamza Mahmood @_hmahmood]

This is meeting rooms II problem on leetcode.

[spidey @lochan_twt]

i got rejected in an interview because of this question

[Hamza Mahmood @_hmahmood]

2022 is the new Year 0 of civilization, from now on, BC doesn’t mean Before Christ, it means Before ChatGPT

#AI

[Hamza Mahmood @_hmahmood]

Say no more

[Dave Griffiths @Dave__Griffiths]

@DeryaTR_ What if they invent their own language which we don’t understand?

[Neil Zeghidour @neilzegh]

Me defending my O(n^3) solution to the coding interviewer.

[Tanishq Mathew Abraham, Ph.D. @iScienceLuvr]

Hot take: too many talented AI researchers are focusing on code generation. There are many other interesting applications of AI that deserve more top talent.

[« 2k2 » @peach2k2]

wild c feature none of you knew about

[Hamza Mahmood @_hmahmood]

@rasmalai Assembly

[Hamza Mahmood @_hmahmood]

Why not let the best coding model have a look?

[Thariq @trq212]

We've received some feedback about a potential degradation of Opus 4.5 specifically in Claude Code.

We're taking this seriously: we're going through every line of code changed and monitoring closely.

In the meantime please submit any transcripts with issues through /feedback

[Hamza Mahmood @_hmahmood]

@Davidrejll @TrevesStudios @drakebonergif Generative Pretrained Transformer.