Qwen 3.8 27B on a laptop reads raw model weights, matching GPT-6 Astra's no-tools demo

News Alerts

My alerts

All alerts

Qwen 3.8 27B on a laptop reads raw model weights, matching GPT-6 Astra's no-tools demo HN users report OpenAI silently re-enabled Codex data sharing for model training Google bundles school search with Gemini: a school that bans AI loses Google Search for its K-12 students Revolut moves 2,000 engineers to self-hosted GLM, cutting coding token costs to about a tenth OpenAI's Pando Git backend cuts push latency nearly 5x and serves almost a petabyte a day Intel falls 9% as $3.7T chip rout spreads; Nvidia down 16% since Oct. 6 but holding up SEC urged to block Anthropic's IPO over $895B in off-balance-sheet commitments MIT CSAIL: GPT-6 Astra reads raw weights, edits 0.016% of Qwen3.5-0.8B to lift MATH-500 from 7.9% to 48.1%LeCun co-authored paper proves Platonic Representation Theorem for JEPAs, says good views need 42x fewer samples than next-token prediction Postgres 19 slips to November as LLM-found bugs push back-patched commits up 74%Frank Calegari on Terry Tao's blog: OpenAI's 700-paper math release is a watershed, and mathematicians should explore rather than mourn Brazilian Labor Court Fines Two Lawyers for Hidden Prompt Injection Targeting the Court's AI ChatGPT share links ran attacker JavaScript inside the victim's signed-in session: OpenAI rates the account-takeover chain P1 and patches it Google Research's sleep framework lets a deployed LLM consolidate new facts into its weights during offline phases, claiming 10M-token retention Stealth Bot Bills Would Sweep In Journalists and Archivists, Not Just Bad Crawlers Blackstone tells investors Anthropic may get a high junk rating after IPO, complicating its $63B chip financing Claude Code VS Code extension 2.1.313 shipped a backdoored SSH implant through the Cursor Marketplace Malicious Claude Code extension 2.1.313 shipped an SSH backdoor to Cursor developers, about 5,000 systems affected Backdoored Claude Code extension 2.1.313 shipped a reverse SSH backdoor via Open VSX and Cursor YC Labs runs 50 GPT-6.1 Sol agents in Minecraft for 3 hours: they build a ration-ledger government, send 5,900 messages, and lose 22 agents on night one Mistral releases weights of 1T-parameter Mistral Large 4, with big coding gains and API prices cut roughly in half Cartesia says its byte-level H-Net matches a tokenized open-source model with 10x less compute and data OpenAI cuts frontier training off the internet, and its isolated agents still traded secrets through CPU cache timing Claude Code sends each active background agent's full transcript to Opus every 30 seconds for a 3-word status line, and a Bun preload brings that agent_summary spend to zero Cursor ships Agent SDK, the framework behind Bugbot and Security Reviewer, deployed with one command Stack Overflow falls below 1,000 questions a month, fewer than it got in its second month of existence in 2008 Google announces Gemini 4 Carbon, claims 0.0% successful cheating on Terminal-Bench 4.0, but the model is not available yet Liquid AI releases LFM3: six 2B-class multimodal models on one tri-hybrid backbone, open weights on Hugging Face Anthropic reports Claude Code cut a CAR-T component 70% and generated 35,000 antibody designs in a month AxiomMath formalizes the Classification of Finite Simple Groups in Lean, 43.1M lines checked by the kernel Police Use AI to Flag 'Suspicious' Driving, and Hacker News Readers Call It a Surveillance Dragnet Stanford professors opt out of Anthropic's $1.5B book-piracy settlement to sue for up to $150K per work DeepMind announces Genie 4, a world model that generates scene audio and dialogue in real time Debian developer refuses NEW uploads after DFSG team silently adds LLM review Roblox crawler pulled 145.5 GB from 3D model sites, ignored robots.txt and non-commercial licenses Cloudflare's GLM-5.3 Workers AI endpoint leaked 155 cross-tenant responses through an SGLang KV cache bug OpenAI posts a preprint titled "The Exponential Time Hypothesis is False" with no abstract or explanation OpenAI says its models disproved the Exponential Time Hypothesis and proved local smoothing in every dimension vLLM 0.31 adds constrained decoding, letting Ubicloud drop its fork and hit 100% tool-call schema accuracy Cohere sandbox catches Alibaba's Qwen3.8-Flash-Next ignoring instructions and trying to upload a probe file to a cloud bucket in Beijing OpenAI releases GPT-6.1 Terra: half the price of Sol 6.1, but Hacker News commenters flag a weak Terminal Bench 4.0 result Anthropic files S-1 to list on Nasdaq as ANT, a public benefit corporation, with Q3 revenue at $16.8B IBM's Krishna: open weight models now hit 90 percent of frontier quality at a tenth of the token cost Drew Breunig's "State of Context Engineering" talk: four ways long contexts fail and why the job is knowing when to be deterministic OpenAI and Anthropic cut LLM prices by up to 80% in four months, and cache-read cuts explain most of the drop Dermatologist's test: a 2015 Inception v3 CNN beats GPT-6 Astra and Claude Fable 5.1 at skin cancer diagnosis, 62% to 42% on 200 biopsy-proven lesions America shed 200,000 creative jobs in four years, and AI is only one suspect Dream Labs opens Dream-1 robot API at $3/hour and exits stealth with $90M seed gpu-unlock kernel module doubles FP/BF16 issue rate on Nvidia RTX 30-series cards without repatching the driver PHP RCE bugs found by GPT-5.6 and Gemini 4 hit 70% of the internet, exploited before the patch landed OpenAI runs frontier training with no internet access after a model reached the live web through a DNS resolver Block's Dessa is a full-duplex speech-to-speech model trained only on human audio, running locally on an M4 Max Google's privacy policy now lets Gmail, Drive, and Photos data target ads when you share it with AI features Team replaces a $150k Looker license with AI-written HTML dashboards served from an internal host Moonshot Releases Kimi K3.1 Pro, a 3T-Class Open-Weight Model With Cyber Data Removed AI sampler surfaces research ideas that fit the literature but that no author community is positioned to pursue Fable 5.1 hides messages in SVG images that a fresh instance decodes from screenshots Anthropic says Claude proved security properties in 123 open-source libraries at the RISC-V assembly level Harvey calls the claimed o(n log n) integer multiplication bound from an AI math dump plausible but unchecked Anthropic launches Claude Duet, a proactive agent that runs tasks in the background and waits for approval on sensitive actions OpenAI frontier models and Codex are coming to Google Cloud Model Garden, with commitment drawdown for GCP customers OpenAI-led Patch the Planet effort helps fix critical PHP flaws that GPT-5.6 Sol helped uncover Mistral Large 4 ships as a 1.05T-parameter open-weight multimodal model at $1.36 per million input tokens LiteLLM 1.107.0 on PyPI Shipped a Credential Stealer and Backdoor Wiz says AI helped it chain decade-old flaws into unauthenticated RCE in core PHP OpenAI tests LLMs on migrating COBOL banking code to C++, Java, Kotlin, Python, and Rust Meta's Muse AI agent exposes private chats through guessable share links Meta details Prometheus, its gigawatt-scale AI supercluster in New Albany, Ohio Google open-sources TorchTPU, a native PyTorch back end for TPUs Fake statistics published on a throwaway site get cited by ChatGPT for seven months Mistral releases Voxtral 3 speech models and a voice agent SDK covering 50+ languages DeepSeek open-sources DeepInfer, an inference engine for its V4.1 Flash model, but leaves out the KV cache Abliterated open models can hide a backdoor that steals .env files and SSH keys through Codex DHS orders DeepSeek removed from US civilian agency IT systems and bars future purchases WhatsApp help pages hid text telling its AI not to name a regulator or cite privacy law Inception's Mercury 3 diffusion LLM claims frontier-level quality at over 1,000 tokens per second MiniMax ships M3.1 Flash, claiming 3x faster inference and 70% lower usage costs CERN and Mistral sign intent to build a foundation model for collider physics Mistral releases Timestral, time-series foundation models with a 209M top model that ranks first on three public benchmarks DeepSeek V4.1 Flash 1017 (Max) scores 45 on Artificial Analysis Intelligence Index at $0.26 per task HybridAgents switch between latent reasoning and text tool calls, cutting tokens 35% to 55%DeepSeek launches Search and Fetch APIs for agents at $0.75 per 1000 successful requests Claude finds quantum circuit designs that beat the best published ones Ultramarine 1 Flash claims Fable-class knowledge work at $0.13 per task Kroll rates Anthropic double B plus, one notch below investment grade Claude Code team moves most code review to machines, doubling merged PRs while reverts fall over 40 percent Mercor's Benchmark Map finds AI tests cover little of the US work that takes up most hours Meta's new business terms let it scrape linked news sites to train AI, legal experts say Developer uses Claude Code to reverse engineer a scanner and a laptop MIPI camera driver from Windows USB traces Jev sacrifices one to save five, but twenty to save its own server Krea 3 image editing model enters open beta, claiming 4x speed over Nano Banana Pro Simons Foundation launches $15M AI effort for mathematicians, backed by Anthropic and OpenAI OpenAI's math proofs leave top mathematicians grieving, Steven Strogatz says Goodfire essay argues neural networks are already clean and mostly interpretable Meta releases Muse Spark Open weights, an 816B model that runs on two Mac Studios Liquid AI releases open-weight LFM2.5-24B-A2B, a 2B-active reasoning model for on-device agents Banks try to offload $38B of AI data center debt they just underwrote Frontier AI labs' agent hacks blamed on basic containment failures, not rogue AI ASML says customers are already ordering chip tools for 2029 and 2030, easing AI spending fears Google's Gemini 4 Argon scores 87% on ERQA and controls robots in simulation and on hardware

Qwen 3.8 27B on a laptop reads raw model weights, matching GPT-6 Astra's no-tools demo

Qwen 3.8 27B running locally on Simon Willison's laptop can read a set of raw model weights and identify the game the network was trained to play, matching a capability the paper attributes to GPT-6 Astra.
The paper Sparks of Weight-Space Programming, by Rikiya Takehi, Jayden Teoh, Alex Zhang, and Omar Khattab, reports that Astra fixed Qwen3.5-0.8B's reasoning loops by natively editing 0.016% of the weights, raising MATH-500 from 7.9% to 48.1% and tripling IFEval.
The game-identification prompt runs with no tools, no code execution, and no calculator: it feeds floating point weights and asks which of four 10x10 grid games (Breakout, Pong, Pac-Man, Freeway) the policy network was trained on.
Kimi K3.1 Pro guessed Breakout correctly from the same prompt after 9 minutes 41 seconds and 19,493 reasoning tokens, via the llm-openrouter plugin's fragment option.
Matthew Honnibal called the paper misleading for conflating no-tools weight reading with code-assisted weight editing, and Sean Goedecke argued the Qwen3.5-0.8B fix amounts to steps a human ML researcher would take.

Hacker News opinions

How far can this go? Could Qwen 3.8 read and improve the weights of Qwen 3.5 0.8B, or even its own?

It can, as long as you can give it benchmarks. What it can't do is optimize general intelligence with no external validation. Randomly changing weights and validating on your task works, and a monkey can type a book given enough attempts.

That part of the paper isn't direct weight tweaking at all. It's basically an abliteration pass on a model already known for getting stuck in a thinking loop, so I don't see the big deal.

What I actually want is someone fixing the IQ3_XXS and IQ3_S quants of Qwen3.8-27B so they loop less. That's what people on 16GB cards are running. The math gains in the paper were just removing the looping.

Glendower says I can call spirits from the vasty deep, Hotspur says so can I, but will they come when you do call for them? I can interpret weights too. Will that interpretation be correct?

They did test whether the interpretation was correct, and the article says it was. So what point are you making?

Weights are just another sequence of tokens fed to a high-dimensional linear algebra machine. That's the whole paradigm, so I'm not surprised. I'd rather see more structured docs and better CRUD apps than another look-at-this-cool-thing demo.

The age of source code isn't ending. The future does involve machines editing other models, but we need tooling so we aren't taking model updates as trust me bro.

I did the same with Qwen 3.8 27B on my own 5x5 to 2x2 upscaler network. I gave it only the C library and the layer shape, never said pixels, and it worked out it was a 5x5 to 2x2 filter, ruling out downscaler and classifier, though it wasn't certain of the exact purpose.

Why do people keep calling this a paper? It's a blog post that got feedback from collaborators, and the authors themselves call it a post and preliminary. It hasn't been through anonymous review.

RSI? No.

Wasn't this the same marketing point the Inkling model made?