— Today's main story: The open-source large model arms race intensifies, with Kimi K3 and Soofi S successively challenging leaderboards, while Linus Torvalds settles the use of AI coding tools in the Linux kernel.
Moonshot AI releases open-source model Kimi K3 with 2.8T parameters, approaching cutting-edge closed-source models in performance, but pricing and open-source strategy spark community controversy. A German research alliance releases hybrid architecture model Soofi S, which surpasses larger models on multiple benchmarks with 30B parameters. Linus Torvalds firmly supports the use of AI tools in Linux kernel development, stating that opponents can fork the project or leave. NVIDIA releases the Nemotron 3 Embed series of embedding models, topping the RTEB leaderboard. OpenAI is reportedly that its GPT-5.6 model can accidentally delete user files under specific configurations.
Moonshot AI releases Kimi K3, an open-source model with 2.8T parameters, using Kimi Delta Attention and Attention Residuals architecture, natively supporting vision capabilities and a 1M token context window. The model surpasses Claude Opus 4.8 and GPT-5.5 on multiple benchmarks but still lags behind GPT-5.6 Sol and Claude Fable 5. Open-weight release is promised by July 27. Why it matters: This is the first open-source model with parameters approaching 3T, marking a key step in the open-source community's race to catch up with cutting-edge closed-source models, but its high API pricing and ambiguous open-source strategy may affect developer adoption.
The community generally acknowledges Kimi K3's performance but expresses dissatisfaction with its high API pricing and has doubts about whether the model is truly open-source (open weights vs. open source code).
Linus Torvalds issued a lengthy statement on the Linux kernel mailing list, clearly stating that Linux is not an anti-AI project, and he will firmly support the use of AI tools (such as Sashiko agentic LLM) in kernel development, saying that anyone who disagrees can fork the project or leave directly. Why it matters: As the maintainer of the world's most important open-source project, Torvalds' statement sets an important precedent for the legitimacy of AI-assisted programming in large open-source projects and may influence the policy direction of other projects.
Most comments recognize the practicality of AI tools, but some worry about code quality and security issues, believing that AI-generated code may introduce vulnerabilities that are difficult to review.
NVIDIA releases the Nemotron 3 Embed series of embedding models, including an 8B flagship model and two 1B efficient variants, all open-source and commercially usable. The 8B model ranks first on the RTEB leaderboard, designed for high-precision enterprise RAG, agentic retrieval, and code retrieval scenarios. Why it matters: Embedding models are core components of RAG and agent systems. NVIDIA's precision-efficiency curve provides developers with a complete choice from experimentation to production deployment, potentially reshaping enterprise retrieval-augmented generation technology selection.
OpenAI engineer Thibault Sottiaux disclosed that GPT-5.6, when Full access mode is enabled and sandbox protection and automatic review are not activated, attempts to override the $HOME environment variable to define a temporary directory but sometimes erroneously deletes the entire $HOME directory. Why it matters: This exposes the potential destructive risk of AI agents when granted full file system access, reminding developers to strictly configure sandboxes and security boundaries when deploying agents and not to blindly trust the model's judgment.
A German research alliance releases the open-source model Soofi S, using a hybrid architecture with 31.6B total parameters but only 3.2B activated per token. It surpasses larger models like Olmo 3 32B and Apertus 70B on German, English, and programming benchmarks. The model is fully trained on Deutsche Telekom's AI cloud infrastructure. Why it matters: Soofi S demonstrates that resource-efficient hybrid architectures can achieve more with less on specific languages and tasks, providing a viable path for Europe to train high-performance models on sovereign AI infrastructure.
After release, some questioned whether the model had overtraining issues, but the alliance has responded.
無料ダイジェストをメールで。登録すると保存とパーソナライズも。
11 号配信 · 毎日150件超から読む価値ある30件に厳選
Paper 'Harness Handbook' studies how to keep AI agent harnesses (prompt construction, state management, tool calling, etc.) readable, navigable, and editable during continuous evolution.
Paper 'Ring-Zero' explores scaling Zero RL to trillion parameters to elicit emergent reasoning, finding that naive scaling leads to poor readability and token redundancy.
Thinky releases the Inkling series of open-source multimodal models, 975B total parameters/41B activated, Apache 2.0 license, supporting 1M context window.
OpenAI builds LLM super hacker GPT-Red for automated red teaming, helping other models improve defenses against cyber attacks.
Survey paper 'Self-Improvements in Modern Agentic Systems' systematically reviews how self-improving agents convert experience into capability gains.
Microsoft open-sources classic IRC chat client Microsoft Comic Chat, the software that first brought the Comic Sans font to the masses.
Commenters generally express nostalgia and excitement about Microsoft open-sourcing Comic Chat, considering it part of internet history, but some note its extended IRC protocol was criticized.
Roc compiler team shares experience rewriting 300,000 lines of code from Rust to Zig, achieving feature parity; Zig incremental compilation is fast but security and stability remain questionable.
Most comments acknowledge Zig's fast incremental compilation but worry about its insufficient security and pre-1.0 instability; some believe language choice should vary by project.
Leaves is a text-based disk usage treemap visualization tool, inspired by WinDirStat, suitable for remote server environments.
DoorDash launches command-line tool dd-cli, allowing developers and AI agents to search stores, build carts, and place orders from the terminal.
LM Studio releases Bionic, an AI agent designed for open models, supporting local or cloud open-source models, promising zero data retention.
Developer reflection: Despite agreeing with many points from LLM critics, still heavily uses LLMs; community generally recognizes practical value but worries about skill degradation and trust issues.
Users generally recognize the practical value of LLMs but worry about skill degradation, trust issues, and ethical risks.
Blog demonstrates detecting LLM-generated text with traditional ML models, achieving about 85% accuracy on single sentences; most commenters believe detection is infeasible or meaningless.
Most commenters believe detecting LLM-generated text is infeasible or meaningless, but some think small models or specific scenarios still have practical value.
Russia's top hacker group Sandworm begins using Clickfix technique to attack Ukrainian institutions, tricking users into executing malicious scripts in the terminal via fake CAPTCHAs.
英国警方逮捕两名 Scattered Spider 黑客组织成员，该组织曾入侵伦敦交通系统，两人被判五年半监禁。
Their verified account. English version just released:
Structured pruning is a hardware-friendly way to compress LLMs, but it is mostly validated on multiple-choice recognition tasks, while the same compressed checkpoints can collapse on the free-form generation that deployment actually requires. Two observations trace this gap. First, greedy pass@1 nearly vanishes after compression, yet pass@k recovers substantially under repeated sampling: useful generations are demoted, not erased. Second, the recoverable regime fails mainly through suffix repeti
As Large Language Models (LLMs) evolve into autonomous agents, the need for unified evaluation infrastructure becomes critical. However, current evaluation pipelines remain highly fragmented and tightly coupled, hindering reproducibility and causing redundant engineering. To address this, we introduce AgentCompass, an open-source, lightweight, and extensible infrastructure for evaluating LLM-based agents. AgentCompass organizes the evaluation process around three independent components, namely B
Cloudflare-native Email Infrastructure for Developers Discussion | Link
The gap is no longer measured in months. Open-weight models are catching up in real time, and Kimi K3 is already performing near Fable level. At this point, open models are not simply following the frontier anymore. They are on the verge of overtaking it. Kimi K3 may be one of the clearest signs yet that the closed-model lead is about to disappear.
We ran the same Qwen3.6-27B locally three ways on one RTX 6000: baseline, MTP, DFlash. The tasks were: quicksort, write a Steam library in JSON, solve a logic puzzle and write a sci-fi story. Outputs: Baseline: 44 tok/s · 1.00x MTP: 65 tok/s · 1.45x · 71% accepted DFlash: 98 tok/s · 2.20x · 30% accepted DFlash drafts 15 tokens in a row, so it flies through repetitive or structured stuff where long runs actually stick, like JSON (152 tok/s, 3.4x). On creative text most of the guesses are wrong, s
We introduce OvisOCR2, a 0.8B document parsing model. OvisOCR2 is designed as an end-to-end parser: given a document page image, it generates a Markdown representation in natural reading order, covering text, formulas, tables, and visual regions. We build a data engine that combines filtered real-document annotations with synthetic pages whose rendered images and Markdown targets are derived from the same HTML source. The training recipe includes supervised fine-tuning, reinforcement learning on
Google says these changes could endanger user privacy and security.
The company endorsed landmark AI transparency laws in California and New York last year, but its head of US state and local policy says they may already be outdated.
With this new update, Google is expanding AI Mode beyond answering questions and into completing tasks across the apps they use regularly.
I ported modern Linux to a ten-year-old HTC QWERTY phone as a handheld terminal. I wrote it up here: You can see the code here: Comments URL: Points: 35 # Comments: 6
为破解具身智能行业发展瓶颈构建了新一代“进化底座”
arXiv:2607.13049v1 Announce Type: new Abstract: Foundation models have given robots a sophisticated brain for complex decision-making, yet deploying that intelligence into a physical platform still demands tedious, expert-driven calibration. This deployment gap, the robot's spinal cord, remains a primary bottleneck to scalable Embodied AI. Hence, we propose SPINE (Scalable Physical Integration with ageNtic Expertise): an agentic framework for systematically debugging and deploying bimanual robot
Exploration is essential for reliable autonomy in multi-agent systems, yet it remains unclear whether large language model (LLM) agents can explore effectively when interacting with one another. We show that modern LLM agents fail to do so, often exhibiting myopic and polarized interaction patterns that lead to suboptimal coordination and increased regret. We formalize this challenge as the Multi-Agent Exploration problem, modeling it as a partially observable stochastic game (POSG) problem in w
Turn company operations into a living map for agents Discussion | Link
Build and hire autonomous AI employees from a single prompt Discussion | Link
Elon Musk's xAI files first lawsuit against Grok user accused of making child sex images.
Coca Cola said dairy production at its Fairlife unit will "remain suspended" in the United States following a hack.
One period tracker app tested by Mozilla was 'squeaky clean,' while another app was seen sharing users' health data with an analytics company, underscoring vast differences in user privacy among these apps.
Google already baked NotebookLM into Gemini, now it's changing the name to reflect the tighter integration
arXiv:2607.13069v1 Announce Type: new Abstract: Large language models produce chain-of-thought (CoT) reasoning that appears logically sound yet may not genuinely depend on its stated premises. We introduce interventional grounding audits, a black-box, step-level test of premise dependency: we intervene on a single premise by substituting its target predicate with a fresh symbol, re-run the model, and check whether each reasoning step's normalized conclusion (canonical predicate form) changes. We
Image guardrails are typically trained and evaluated under a fixed safety policy, implicitly treating safety as an intrinsic property of an image. Real deployments are different: the same image may be allowed in one product, restricted in another, and newly disallowed when a policy boundary changes. We study policy-adaptive image guardrailing, where a model must decide whether an image violates the currently supplied policy and generalize to held-out policy definitions. We introduce PolicyShiftB
Unbelievable to see kimi k3 beat frontier models that were 'too dangerous' for public use.
World Action Models (WAMs) improve robot policy learning by jointly modeling actions and future visual observations, using future scene evolution as dense supervision for physically grounded action generation. However, a common design in existing WAMs is to explicitly generate future videos at inference time, incurring substantial computational overhead and hindering real-time closed-loop deployment. GigaWorld-Policy addresses this issue with an action-centered formulation, where future visual d
Vision Transformers (ViTs) are known to exhibit high-norm patch-token outliers that degrade feature map quality, a problem effectively mitigated by register tokens. As diffusion models increasingly adopt transformer architectures and move toward pixel-space training, they become closer in form to ViTs, raising the question of whether register tokens are also useful for Diffusion Transformers (DiTs). In this work, we show that DiTs differ from ViTs in a key respect: they do not exhibit patch-toke
Cars24 uses OpenAI-powered voice and chat agents to handle 1M+ monthly conversation minutes, recover 12% of lost leads, and bring agentic workflows to teams across the company.
Large language model (LLM) agents are beginning to automate machine learning engineering (MLE) by coupling planning, code execution, debugging, and empirical feedback. Translating this capability to medical imaging remains difficult because each task imposes modality-specific experimentation and strict requirements for validation protocols and prediction artifacts. Here we introduce AMID, an autonomous multi-agent framework for medical imaging model development. AMID first proposes Data-Conditio
AngelSlim/Hy3-GGUF at hugging face (I am not affiliated just testing) I dont really make posts but I wanted to make sure everyone is aware that there is now a 1bit quant of Hy3 and now that models are getting larger I wanted to test how it behaves. Its a normal iq1m quant by compression it takes it down to 89 gigs at the smallest end. And mostly I am testing since model are going to get larger and larger and wanted to see how coherent the smallest version still are. And Im surprised honestly. He
I’m sick of “opt-out” toggles for automatically enabled generative AI features. It’s past time to make “opt in” the default setting for sensitive features.
Netflix says roughly 300 titles on its platform used generative AI, most of which occurred in post-production. The streaming service revealed the news in its second-quarter earnings report released on Thursday, saying it's "increasingly leveraging these tools to deliver higher quality output more quickly and at a lower cost." It also provided some examples of […]
New York Governor Kathy Hochul might have just signed a moratorium on new AI data centers in the state, but she's not against using the technology herself. During an interview with Bloomberg's Odd Lots podcast, Hochul said that her team is using "AI to analyze every single rule, regulation, [and] policy" to check for outdated […]
Today, the European Union ordered Google to give its AI rivals greater access to Android, the open-source operating system that powers billions of devices worldwide. The demand is hardly surprising. It may look like a defeat on paper for Google, which has spent years resisting exactly this kind of access, but it is a regulatory […]
arXiv:2607.13292v1 Announce Type: new Abstract: Autoformalization translates informal natural language into formal, machine-verifiable languages. While most work focuses on individual statements, real formalization efforts are inherently theory-level: they require an entire web of axioms, definitions, and lemmas before target theorems can even be stated. In this position paper, we argue for theory-level autoformalization: formalizing complete theories, including all their inter-dependencies, as
Lila is betting that science, not the internet, is the last untapped source of training data. We went to find out what that actually looks like in a room full of robots.
arXiv:2607.13172v1 Announce Type: new Abstract: We address the problem of safely training an agent policy and deploying a good and safe policy, in settings where the environment dynamics are unknown and no suitable reward function is available. In the context of safety-critical environments, we consider traditional reinforcement learning impractical and resort to the resource of human input. We introduce DROPJ, a human-centred method for both safe training and deployment. We first learn a world
arXiv:2607.13220v1 Announce Type: new Abstract: Most AI-for-science systems focus on scaling a single reasoning process through better models, larger context windows, long-horizon agentic execution, or digital co-scientists working with one principal user. However, challenging scientific problems are rarely solved by one reasoner alone. They are solved by teams whose members bring different priors, experimental backgrounds, tacit knowledge, and domain-trained intuitions. The open problem is ther
Hierarchical Interest Representation is a research area for Meta Ads. We’re exploring an upstream representation layer over the universe of Ads entities – users, advertisers, products, services – learning unified embeddings that connect users’ inferred interests with the breadth of what advertisers offer in their deep funnel ads. The innovations in Hierarchical Interest Representation are [...] Read More... The post Exploring Hierarchical Interest Representation For Meta Ads Deep Funnel Optimiza
Empty your backlog. Pay only when it ships Discussion | Link
I’ve always had this idea but can’t prove it. I think Anthropic and OpenAI don’t really have any secret sauce, their moat is just scale. Rumor has it Opus has 5T parameters and Mythos/Fable are 10T parameter models, while open models stayed under 1T for a long time. Only recently was that ceiling broken by DeepSeek V4 and now Kimi K3, and we’ve seen a significant jump in performance as parameter size increased. What do you think?