Solal (@ArxSys) profile picture

Security researcher and lead dev. of TAP-IR & DFF (open source softwares for incident reponse). If you like retrogaming & fpga dev follow my account @s_on_vg

Paris
Joined July 2010
Photo by Solal (@ArxSys)Photo by Solal (@ArxSys)Photo by Solal (@ArxSys)Photo by Solal (@ArxSys)Photo by Solal (@ArxSys)Photo by Solal (@ArxSys)Photo by Solal (@ArxSys)Photo by Solal (@ArxSys)Photo by Solal (@ArxSys)Photo by Solal (@ArxSys)
Hackers are now using LLMs to continuously scan software updates, spot newly fixed bugs, and build attacks before teams can apply the fix — all without human involvement. That’s according to the latest report from the Google Threat Intelligence Group. We're helping defenders counter automated threats with Google AI Threat Defense, which combines Gemini and other frontier models to find and fix flaws before attackers can exploit them.
765
6,269
1,773
38,187
4,674,824
Can you design a chip? Jane Street's new competition: design an open-source, reprogrammable protocol emulator ASIC. The most novel designs get made on our March 2027 IHP shuttle! 🚀 Closes January 18th: blog.janestreet.com/protocol… #ASIC #OpenSource
2
23
3
200
7,747
BREAKING : Golden Owl a recensé 164 comptes opérés depuis l'Iran qui s’adressent à un public français sur X, dissimulés derrière des identités fictives ; leurs publications ont généré au moins 46 millions de vues et 1 081 611 interactions. Ces faux comptes se sont immiscés pendant trois ans dans les mouvements de contestation en France ; ils ont repris leurs activités à l’automne 2026 et exploitent les troubles sociaux en ce moment même. goldenowl.ai/resources/resea…
20
387
18
572
14,883
(1/5) First a CPU, then a GPU, then a filesystem. Next: a research lab. Meet qwenLab: Qwen3.5-9B with six small transformers living inside it. Each is a 420k-parameter model learning modular addition, the classic grokking task. The model trains them side by side in its own forward passes with Qwen deciding how.
First a CPU, then a GPU. A computer isn’t complete without storage. Meet qwenFS: Qwen3.5-2B with a filesystem compiled directly into its weights. Nothing runs outside the model. It adds 8 layers in front of Qwen’s own 24, using attention to hold the files and Gated DeltaNet state to hold the tables. Every save and read runs in the model’s own forward pass. A write stores its text in the copy layer’s attention log, keyed by file, version and offset. A read uses those same keys, so attention copies the text back exactly token by token.
9
18
1
157
5,016
Inspired by QwenGPU, I built QLink with Opus 5.5. QwenGPU compiles a small GPU into the weights of Qwen3.5-2B, so that the model runs CUDA kernels by decoding tokens. QLink connects several of these models with the protocols real GPUs use and to train neural networks on them.
7
13
148
7,742
Solal retweeted
Meet qwenGPU: Qwen3.5-2B with a SIMT GPU inside its weights, 45 compiled layers in front of Qwen’s own, using its Gated DeltaNet layers as registers and memory. It runs real CUDA in its forward passes. The host compiles the kernel to PTX and loads it, then the model writes every step of the launch itself: every register, branch and memory write. The GPU has 4-lane warps, a 4-stage pipeline, 15 registers per lane and 126 words of global memory, all held in DeltaNet state. Qwen generates each launch token by token, each record running one machine block for one warp.
44
96
32
1,046
152,017
Solal retweeted
First a CPU, then a GPU. A computer isn’t complete without storage. Meet qwenFS: Qwen3.5-2B with a filesystem compiled directly into its weights. Nothing runs outside the model. It adds 8 layers in front of Qwen’s own 24, using attention to hold the files and Gated DeltaNet state to hold the tables. Every save and read runs in the model’s own forward pass. A write stores its text in the copy layer’s attention log, keyed by file, version and offset. A read uses those same keys, so attention copies the text back exactly token by token.
Meet qwenGPU: Qwen3.5-2B with a SIMT GPU inside its weights, 45 compiled layers in front of Qwen’s own, using its Gated DeltaNet layers as registers and memory. It runs real CUDA in its forward passes. The host compiles the kernel to PTX and loads it, then the model writes every step of the launch itself: every register, branch and memory write. The GPU has 4-lane warps, a 4-stage pipeline, 15 registers per lane and 126 words of global memory, all held in DeltaNet state. Qwen generates each launch token by token, each record running one machine block for one warp.
84
153
46
1,828
136,716
Solal retweeted
if the MIT post about Astra capabilities got you interested in weight editing, i have an old post from 2023 about doing exactly this to make a transformer from scratch for a simple sequence prediction task. vgel.me/posts/handmade-trans…
New blog post: making a transformer by hand, without training! Want to understand transformers and attention better? This post goes through assigning each weight for a GPT-2-like transformer to understand how they work.
6
40
1
424
10,623
This is an old hat. That's exactly what the fast weight programmers did in the 1990s (including the 1991 unnormalised linear Transformer). Overview with links to the original papers: people.idsia.ch/~juergen/fas…
15
24
11
391
33,496
This is an old hat. That's exactly what the neural fast weight programmers did in the 1990s (including the 1991 unnormalised linear Transformer). Overview with links to the original papers: people.idsia.ch/~juergen/fas…
We asked GPT-6 Astra to edit Qwen3.5-0.8B weights to improve math. It directly changed only 0.016% of weights, fixing 0.8B's reasoning loops & improving math up to 6x and even instruction following 3x, without gradients or training. We call this Weight-Space Programming: frontier LLMs can now natively read/write neural network weights! 🧵 Blog: rikiyat.github.io/blog/weigh…
Readers added context they thought people might want to know
LLM was given weights explicitly as text in prompt and used code/forward passes (per authors' logs) to ablate Qwen looping direction, a known thinking-mode issue per model card. Primarily fixed repetition in that mode, not general 'natively' improving weights/math. rikiyat.github.io/blog/weight-sp… github.com/RikiyaT/lm_und… huggingface.co/Qwen/Qwen3.5-0… huggingface.co/RikiyaT/Qwen3.…
101
199
99
2,104
231,923
We asked GPT-6 Astra to edit Qwen3.5-0.8B weights to improve math. It directly changed only 0.016% of weights, fixing 0.8B's reasoning loops & improving math up to 6x and even instruction following 3x, without gradients or training. We call this Weight-Space Programming: frontier LLMs can now natively read/write neural network weights! 🧵 Blog: rikiyat.github.io/blog/weigh…
Readers added context they thought people might want to know
LLM was given weights explicitly as text in prompt and used code/forward passes (per authors' logs) to ablate Qwen looping direction, a known thinking-mode issue per model card. Primarily fixed repetition in that mode, not general 'natively' improving weights/math. rikiyat.github.io/blog/weight-sp… github.com/RikiyaT/lm_und… huggingface.co/Qwen/Qwen3.5-0… huggingface.co/RikiyaT/Qwen3.…
2,366
7,256
5,753
55,937
10,071,466
BREAKING: Microsoft, $MSFT, is offering buyouts to 60,000 workers as part of the company’s efforts to cut costs amid AI adoption, per Bloomberg. Personnel on select teams in the US, Canada, India, and the UK at a vice president level and below will be eligible. This reflects ~27% of Microsoft's total workforce.
991
2,179
1,787
20,411
11,919,447
‼️ BREAKING: The private keys of 8 certificates from Indian certificate authority eMudhra were compromised, Mozilla says. Firefox blocked them within 3 hours. Each key should be a 116-digit number, but one was just 4. Anyone can guess them and forge HTTPS certs for any website.
618
3,695
1,718
30,104
5,828,800
Transition from vulpocalypse to breachpocalypse is here. IR firms are saturated & turning away work. Now one hacker unleashed AI agents on 1,200+ websites, compromising 100+ for ~$5,000 in tokens. $50 per hack. Agents won and compromised targets. reuters.com/legal/government…
19
63
23
237
56,022
Some magical progress towards "magic state" in quantum! Our @GoogleQuantumAI team has demonstrated "magic state cultivation" on physical hardware for the first time. The result is a 40x error reduction on our Willow processor vs. previous methods, achieving a state-of-the-art 99.99% logical fidelity. It's another step towards useful, error-corrected quantum computers to help solve the hardest problems. The result is published in Science today, watch the video below to see how it works
421
1,104
232
9,271
697,354
Un piratage massif sur des équipements professionnels de réfrigération serait en cours.
Piratage massif sur la régulation Danfoss : cela veut dire que bcp de magasins, entrepôts et restaurants ont la régulation de leurs frigos en croix. Chez moi, c'est un début d'incendie maîtrisé. Les consignes ont été modifiées à 50°.
34
136
14
372
100,449
Replying to @kmad @matthew_pines
I think we might lose public key cryptography.
242
428
457
5,326
2,023,504
So I got access to abliterated GLM 5.3 and either it’s overconfident or we’re all screwed.
44
51
13
1,437
119,882
Imagine the amount of Java crashes this poor LLM must have in its training set that it decided 'NullPointerException' deserves its own token
54
306
26
10,587
311,632