— The open-source model arms race is heating up, and the security response tempo has been completely rewritten by AI.
GLM-5.3 tops multiple coding and security benchmarks with open weights, released the same day as Qwen3.8-Flash-Next with highly convergent architecture; Anthropic releases MHS hardware standard, Claude begins directly controlling physical devices like robotic arms; OCaml maintainer finds that attackers can probe within minutes of a patch being publicized based on mere vulnerability rumors; Vercel open-sources WebGPU library vgpu, turning .wgsl files into importable TypeScript modules.
Z.ai releases GLM-5.3, sharing the same base model as GLM-5.2 with all gains coming from post-training; it improves 50% over GLM-5.2 on Z.ai Code Bench and achieves open-source SOTA on Terminal Bench 3.0, Agents' Last Exam, and CyberGym vulnerability discovery, with exploit-chain benchmark scores doubling. Why it matters: Open-source models are for the first time systematically approaching top closed-source models on complex coding and long-horizon tasks, and the post-training-driven capability leap provides a new upper-bound reference for inference optimization and agent workloads.
Commenters generally consider the performance excellent and cost-effective, close to top closed-source models, though some think the price is high and certain features are missing.
Anthropic introduces the Model Hardware Standard (MHS), translating control methods for hardware from different vendors into a consistent interface that agents can discover, read, and invoke; Claude can now directly control the SO-ARM101 robotic arm in the Hugging Face LeRobot ecosystem without pretrained robot policies or teleoperation. Why it matters: This is essentially MCP for the physical world, extending agents from pure software tool invocation to distributed embodied control, meaning a new integration paradigm for developers in robotics, lab automation, and similar domains.
OCaml core maintainer Anil Madhavapeddy, while fixing the cohttp 6.3.0 path traversal vulnerability, found that his server received probes targeting the vulnerability pattern about ten minutes after the public PR; he himself could also find an exploit using a coding agent based on only a rough vulnerability description. Why it matters: Modern coding agents can turn the faintest vulnerability leads into executable exploits, and the traditional open-source flow of 'fix privately first, then disclose publicly' is no longer sufficient—the security response tempo needs fundamental change.
Vercel open-sources vgpu, a TypeScript WebGPU library that treats .wgsl shader files as importable modules, allowing the same shader to run in the browser, in headless Node.js via Dawn, and in deterministic CI mocks, with full-screen effects at only 25 KB gzipped. Why it matters: WebGPU adapters, bind group layouts, and pipeline descriptors have been the biggest obstacle for frontend teams shipping shaders, and vgpu, released to npm under the MIT license, provides a low-friction engineering path for AI agent shaders and real-time graphics.
Pydantic takes over maintenance of HTTPX, releasing the next-generation Python HTTP client under the name HTTPX2, supporting HTTP/1.1 and HTTP/2 with both sync and async APIs; the OpenAI Python SDK now uses HTTPX2 by default, pulled in automatically when installing openai. Why it matters: HTTPX has long been in a low-activity state yet sits on the critical path of many production systems; Pydantic taking over with a commitment to timely security updates is an important supply-chain stability signal for Python developers relying on the OpenAI SDK or building their own HTTP clients.
The web shows the big picture; subscribers get their own — AI curated to your interests, your private RSS folded in, with community takes, delivered each morning. Free forever.
58 issues shipped · 150+ items sifted to 30 worth reading, every day
The paper proposes using game development as a verifiable trajectory data engine, leveraging compilers and runtimes to provide grounded reward signals for world model RL post-training.
🤖Game engines provide executable verification and long-horizon trajectories for reinforcement learning post-training of spatial world models, motivating a human-engine verification paradigm.
Self-OPD proposes a teacher-free policy distillation framework, using self-explored stochastic branches and normalized advantage optimization for flow matching velocity fields, avoiding the high cost of task-specific teacher models.
🤖Self-OPD eliminates task-specific teachers in flow matching by using self-explored stochastic branches and normalized advantages to optimize the velocity field for multi-objective alignment.
GLM-5.3-Flash and Qwen3.8-Flash-Next independently converged on nearly identical architecture: 3:1 linear mixed attention, 2048-token compression indexer, 4-way gated residual, and Muon optimizer.
Anthropic researcher paper shows an automated system can improve performance on all 10 alignment benchmarks without harming overall performance, providing early empirical evidence for self-improving AI.
GameWAM is the first World-Action Model for native video-game closed-loop control, using block-causal flow matching to jointly predict future visuals and executable keyboard-mouse actions.
🤖GameWAM is a unified world-action model for native video-game control that jointly predicts future visuals and executable keyboard-mouse actions using block-causal flow matching, mode-specific distributions, and block-cycle replanning.
htmx 4.0.0 released, migrating internally from XMLHttpRequest to the fetch() API, developed over 8 months.
Opslane open-sources an agent that records user sessions, identifies user-visible bugs, and only creates PRs when it can verify the fix.
Sesame is a local-first open-source password manager; vaults stay local by default and hosted services never touch the vault itself, licensed under AGPL-3.0.
The author accidentally turned the memory mechanism in an LLM vulnerability research agent into program analysis: the model gradually loses confirmed facts over hours of investigation, requiring structured memory to maintain reasoning consistency.
Luanti was removed from Google Play due to an unsubstantiated DMCA notice from Tracer.AI on behalf of Microsoft; commenters generally believe DMCA abuse lacks penalty mechanisms, though some think Luanti is highly similar to Minecraft.
Commenters generally believe DMCA abuse and false complaints lack penalty mechanisms, causing small projects to be removed without cause; but some think Luanti is indeed highly similar to Minecraft, so the complaint is not entirely baseless.
The US sanctioned Italian left-wing tech organization A/I Collective; most commenters believe there is a lack of evidence and that it suppresses free speech, while a minority think the organization's stance is radical.
Commenters generally believe the US sanctions on the Italian left-wing tech organization lack evidence and suppress free speech, though a minority think the organization's stance is radical and its behavior suspicious.
An article arguing GUIs should be fully keyboard-driven sparked discussion; commenters generally support keyboard operation for efficiency and accessibility, though some believe it shouldn't completely replace the mouse.
Commenters generally support GUIs supporting keyboard operation, believing it improves efficiency and accessibility; but some think keyboard-driven operation shouldn't completely replace the mouse and should balance usability with learning cost.
Users complain about Google apps forcing redirects to the app while the web version lacks features; commenters generally dislike forced downloads, though some think the native app experience is better.
Users generally dislike forced app downloads and believe the web version should retain full functionality; but some think the native app experience is better.
A US judge ruled that the Pentagon's blacklisting of Anthropic was unlawful.
A spy satellite simulator in your browser, except the data is real. Live open source spatial intelligence on a photorealistic 3D globe.
Agent skill for beautiful, verifiable architecture, workflow, sequence, data-flow, and lifecycle diagrams—self-contained HTML with motion and crisp export.
Prompt as Code | GPT-Image2 工业级提示词引擎与模板库，470+ 个案例逆向工程，20+ 套工业级模板，并提炼出Skills，持续更新中
Orca is the ADE for working with a fleet of parallel agents. Run any coding agent with your own subscription. Available on desktop and mobile.
I've been running Qwen3-0.6B on the M5Stack LLM-8850 card (Axera AX8850 NPU, 24 TOPS, 8GB LPDDR4x) hosted by a Raspberry Pi 5 — as a llama.cpp backend. The problem: the vendor stack requires converting every model through their compiler, and their closed runtime gets 13.5–14.5 t/s. I wanted llama.cpp to just work: GGUF in, tokens out. What I ended up doing: Reverse-engineered the engine format. The vendor's compiled engines (.axmodel) store weights in a blob called npu_params. I decoded it: int8
Proof is in the method honestly. Closed model labs need to constantly reinvent the wheel to keep lead. Open source has a bunch of independent labs practically working somewhat together. Eventually when everyone is just releasing weights and papers on how they did it the closed source secrets just get overrun by having plenty of very good secret sauces to the public. That and NO DOUBT chinese labs are sharing internal secrets amongst each which explains how when any of them makes a big jump the o
There's a lot of capital pouring into the business of giving models away.
Their last version 7.14 was released just a month ago. llama.cpp PR(waiting for approval) for Version 10.0 Hope this version comes with more boost & improvements.
It's allegedly been running dozens of gigantic generators without a permit.
A federal judge ruled the Trump administration illegally labeled Anthropic a supply-chain risk, handing the AI company a victory as its second Pentagon lawsuit continues in Washington.
It’s Time. We’re in the Endgame now.
Built the latest and I'm getting as much as 220 tokens per second and averaging in the 170s, I can't get over it. If anyone on here is on that project, fuckkkin' chapeau man, really incredible job. I can't believe I was able to like double or more my throughput from llama.cpp This is what I set up: command: > ninfer-serve /models/qwen3_8_27b_nvfp4.ninfer --model-id qwen3.8-27b-nvfp4 --host 0.0.0.0 --max-context 240000 --kv-capacity 240000 --max-concurrency 2 --kv-dtype fp8 --host-kv-mib 16384 --
I've got Qwen3.8-Flash-next running on RTX 3090, Ryzen 9 3950X, a PCIe 3.0 motherboard, and 64GB DDR RAM from 2020. IQ4_XS weights, full kvarn5 context, vision on GPU, experts in host RAM, n-grams on disk. MTP works but actually slows decode down even with 80% draft acceptance, as expected since every rejected token eats into the host RAM bandwidth. I get 160 tok/s prefill 16 tok/s decode , which makes it a decent option whenever I know I'll be AFK for at least a couple of hours, but not usable
Just as new data centers face growing backlash from neighboring communities, the US Environmental Protection Agency (EPA) is about to make it harder for people to weigh in on any pollution those centers create. The EPA plans to toss out a federal rule requiring public notice and an opportunity to comment when certain industrial sites […]
Neocloud Lambda has raised $1B in private debt to buy Nvidia AI chips and lease them to Microsoft. It's the latest in a string of loans, underscoring the high cost of the AI boom.
A recent paper argues that AI is often better at doctoring than doctors. Guess who isn't thrilled.
"At Hot Chips 2026, Micron drew a notable comparison: For the same memory capacity, HBM requires approximately three times the wafer area of DDR5." "When asked whether this ratio would improve with newer generations, the Micron Fellow reportedly explained that it definitely would not get better." "According to the data shown at Hot Chips, an HBM4 die, for example, operates with 256 memory banks, while DDR5 is specified with 32. Additional data paths, the power supply, and the Through-Silicon Via
The art portfolio platform Cara, designed for creators who don’t want their work used to train AI, has been under assault by trolls seizing and publishing its data.
The firm, known for its focus on software, is going to start throwing more money at the hardware behind AI.
Hello everyone! Pangolin 1.22 introduces a new resource type: AI Gateway. These resources are identity-aware proxies in front of both cloud model APIs and self-hosted model servers, so coding agents and AI clients call a Pangolin URL. A gateway resource can be keyless by authenticating via the Pangolin client or keyed by minting virtual API keys. We're also moving SSH, RDP, VNC, and private HTTPS resources from Enterprise to Community Edition. Pangolin is an open-source, identity-aware remote ac
Zero-shot cross-task generalization, where a policy must execute manipulation tasks never seen during training, remains a central challenge in robot learning. In large language models, a novel task can be performed simply by specifying it in the context, without any parameter update. This form of in-context learning (ICL) turns generalization into a problem of task specification. To achieve cross-task generalization, we bring this paradigm to robotic manipulation, and argue that the natural task
arXiv:2608.26109v1 Announce Type: new Abstract: Machine-learning models can predict ICU mortality accurately, but feature-attribution methods alone rarely provide the clinical narrative needed for bedside use. Large language models (LLMs) may bridge this gap, and multi-step agentic pipelines are a plausible extension because they separate data interpretation, guideline checking, and final explanation. This revised feasibility study preserves the original standalone-versus-agentic comparison whil
As a happy user of ds4, I'm very excited about this branch. Ran some prompts and it seems to be working well on my M4 Max 128gb!
Got a desktop and notebook running NixOS. Now I want to get a beefy NAS and run immich, jellyfin, home assistant, some file shares, vaultwarden and papra on it. My plan is to run NixOS on it because i know it and can restore a PC in a few minutes. The named services do exist on NixOS and that’s how I plan to use them. No docker, no Proxmox. Anyone tried it? Anything I need to know? Objections?
Google is now automatically expanding its AI search summaries at the top of the results page for some searches, as reported by Search Engine Roundtable. The change, when it kicks in, pushes the typical list of links from a search much farther down Google's results page; instead of seeing part of an AI Overview with […]
Anthropic refused to support lethal autonomous warfare and mass surveillance.
Modders are trying out an unofficial version of Nvidia's DLSS 5 on Skyrim, Cyberpunk 2077, GTA V, and a bunch of other games after code for the AI upscaling tech appeared in an early-access build of NBA 2K27. Members of the RenoDX modding channel on Discord reportedly found a way to extract the DLSS "Neural […]
The feature, announced this week, allows Brave's users to sign up for websites and other online services without having to share their personal email addresses.
After an affair with a fellow police officer ended, a Georgia cop used Flock to track her movements—and those of a man whose vehicle often showed up near hers, internal investigation records show.
Meta fixes AI glasses to stop recording any time users cover up the safety light.
arXiv:2608.26113v1 Announce Type: new Abstract: We present PICasso, an AI-assisted framework for automated synthesis, verification, and optimization of photonic integrated circuits (PICs) from natural-language specifications. PICasso couples a structured NL -> YAML -> GDS generation pipeline with PDK aware knowledge injection, automated placement and routing, DRC/LVS validation, and SAX-based photonic simulation. To systematically evaluate AI-driven photonic design, we introduce PIC-Set, a bench
Run prompts on every model at once. Score. Version. Ship. Discussion | Link
11 月 20 -21 日，由奇点智能研究院与 CSDN 联合主办的「奇点智能大会北京站」正式举行
arXiv:2608.26107v1 Announce Type: new Abstract: Predicting students' academic risk in online education is crucial for enabling timely interventions that can improve retention and learning outcomes. However, existing models often suffer from limited early detection capability and insufficient interpretability, leading to a "black-box" trust crisis that hinders their adoption in real-world pedagogical settings. To address these challenges, we propose EduRiskX, a neuro-symbolic framework that integ
You can test it out on breezblue's playground or use it locally, its only ~7GB.
Drop your email to save this and train your feed with 👍/👎 — tomorrow’s digest ranks for you. Free forever, unsubscribe in one click.
Sign up free