News
The raw AI news stream — everything we ingest across RSS, Hacker News, arXiv, GitHub, Bluesky, YouTube and more, newest first.
| Source | Age | Title | Summary | Status |
|---|---|---|---|---|
hn hn-ai | 4d | Winudf – DVD / BluRay driver for DOS, Windows 1.x – 3x / 95 / 98 / ME | — | ↗Not briefed |
bluesky bluesky-ai | 4d | 📋 Unharnessed is an open-source coding agent harness with one idea: the model gets one tool, `ipython`, a persistent #Python REPL. Reading… | 📋 Unharnessed is an open-source coding agent harness with one idea: the model gets one tool, `ipython`, a persistent #Python REPL. Re… | ↗Not briefed |
arxiv arxiv-ai | 4d | Fast at the Frontier, Wrong at the Fallback: A Commit-Level Analysis of Threshold Decoding in Diffusion Language Models | Diffusion language models (DLMs) can write several tokens per forward pass. Open-weight DLMs commit, in any order, every masked positi… | ↗Not briefed |
arxiv arxiv-ai | 4d | SpeCal: Spectral Calibration for Unsupervised RLVR | Reinforcement learning with verifiable rewards (RLVR) has substantially improved the reasoning abilities of large language models, but… | ↗Not briefed |
bluesky bluesky-ai | 4d | OpenAI brought Sam Altman back and that really set a terrible precedence and pushed American corporate governance down the slippery slope. | OpenAI brought Sam Altman back and that really set a terrible precedence and pushed American corporate governance down the slippery sl… | ↗Not briefed |
bluesky bluesky-ai | 4d | Die zentrale These: Jo Bhakdi hält SpaceX für potenziell wesentlich wertvoller, als es selbst seine ohnehin ambitionierte Bewertung vermute… | Die zentrale These: Jo Bhakdi hält SpaceX für potenziell wesentlich wertvoller, als es selbst seine ohnehin ambitionierte Bewertung ve… | ↗Not briefed |
bluesky bluesky-ai | 4d | noteもLLMも10年後にはなくなってるかもしれないし、俺も死ぬかもしれないし、消えてなくなる歴史多いよなあ……って最近つくづく思う。 電話ボックスという文化がなくなったように、10年後には「街の本屋で見かけた」っていう表現もなくなっていくだろうし、まあ3-40年経過すると人… | noteもLLMも10年後にはなくなってるかもしれないし、俺も死ぬかもしれないし、消えてなくなる歴史多いよなあ……って最近つくづく思う。 電話ボックスという文化がなくなったように、10年後には「街の本屋で見かけた」っていう表現もなくなっていくだろうし、まあ3-40年経… | ↗Not briefed |
arxiv arxiv-ai | 4d | VIVID: A Visual Harness for Agentic Video Editing | Video editing requires agents to make editing decisions grounded in audiovisual observations and precise temporal localization, and to… | ↗Not briefed |
arxiv arxiv-ai | 4d | Rethinking Supervision Scaling in On-Policy Distillation: From Dense Supervision to Stochastic Horizons | On-policy distillation (OPD) provides dense token-level supervision along student-generated trajectories, yet the learning benefits of… | ↗Not briefed |
arxiv arxiv-ai | 4d | UbiHearo: Bringing Scenario-Aware Sound Guidance to DHH Users with Mobile Agents | For Deaf and Hard-of-Hearing (DHH) users, effective sound awareness goes beyond detecting \emph{what} sound occurs; it requires unders… | ↗Not briefed |
bluesky bluesky-ai | 4d | Democrat Super PACs are really hitting the LLM bots hard | Democrat Super PACs are really hitting the LLM bots hard | ↗Not briefed |
bluesky bluesky-ai | 4d | None of those are things that'd involve LLMs like those by OpenSI, SpaceXSI and Anthropic, of course. Long before the 'AI revolution' medi… | None of those are things that'd involve LLMs like those by OpenSI, SpaceXSI and Anthropic, of course. Long before the 'AI revolution'… | ↗Not briefed |
arxiv arxiv-ai | 4d | LEAF-OPD: Learning to Forget from Privileged Contexts via On-Policy Distillation | LLM unlearning aims to erase targeted knowledge from large language models while preserving their utility. However, many existing meth… | ↗Not briefed |
bluesky bluesky-ai | 4d | This smells fishy. Either Anthropic is trying to wriggle out of its obligations or it simply can't afford to pay for all the material it st… | This smells fishy. Either Anthropic is trying to wriggle out of its obligations or it simply can't afford to pay for all the material… | ↗Not briefed |
bluesky bluesky-ai | 4d | I am starting to think that Anthropic have about sixty grand in their business account and all the rest of the numbers they’re bandying aro… | I am starting to think that Anthropic have about sixty grand in their business account and all the rest of the numbers they’re bandyin… | ↗Not briefed |
rss techmeme | 4d | Zus Health, which is building an AI-native healthcare data platform, raised a $33M Series B led by Drive Capital, bringing its total funding to $100M+ (Heather Landi/Fierce Healthcare) | Heather Landi / Fierce Healthcare: Zus Health, which is building an AI-native healthcare data platform, raised a $33M Series B led by… | ↗Not briefed |
arxiv arxiv-ai | 4d | MASTA: A Feedback-Scheduled Multi-Agent System for End-to-End Tamarin Modeling and Analysis🔧 GPT-5🔧 Qwen 3 | Tamarin-based formal verification of security protocols requires expert-authored multiset rewriting rules and temporal-logic lemmas. T… | ↗Not briefed |
bluesky bluesky-ai | 4d | Neben den Gründen, die F. Aigner nennt, könnte ein LLM in einem Rechenzentrum im All aktuell schlicht nicht neu trainiert werden. Die verfü… | Neben den Gründen, die F. Aigner nennt, könnte ein LLM in einem Rechenzentrum im All aktuell schlicht nicht neu trainiert werden. Die… | ↗Not briefed |
bluesky bluesky-ai | 4d | AI Models Hate Python and Love PHP: We Asked 12 of Them 1,200 Times, by @den.bsky.social (@jetbrains.com): https://blog.jetbrains.com/ai/2… | AI Models Hate Python and Love PHP: We Asked 12 of Them 1,200 Times, by @den.bsky.social (@jetbrains.com): https://blog.jetbrains.com… | ↗Not briefed |
arxiv arxiv-ai | 4d | Conditional Diffusion Models for Amortized Bayesian System Identification | Generative diffusion models rapidly solve Bayesian inverse problems, such as image restoration and inpainting, with a single trained m… | ↗Not briefed |
hn hn-ai | 4d | Show HN: CFP Tracker – Track Conference CFPs in Obsidian | — | ↗Not briefed |
arxiv arxiv-ai | 4d | ParGLU: Partial Gating Improves Language Modeling in MoE Transformers | Multiplicative gating improves Transformer feed-forward networks (FFNs), but it reduces intermediate representation width under a fixe… | ↗Not briefed |
arxiv arxiv-ai | 4d | Enhancing Mixed Oil Length Prediction through Graph Representation Learning | During the pipeline transportation of refined oil products, the inevitable formation of mixed oil segments poses considerable economic… | ↗Not briefed |
bluesky bluesky-ai | 4d | Overigens wel hardop lachen, hier. (Je moet wat in het zicht van zoveel onbegrijpelijkheid). | Overigens wel hardop lachen, hier. (Je moet wat in het zicht van zoveel onbegrijpelijkheid). | ↗Not briefed |
bluesky bluesky-ai | 4d | Als een sciencefictionauteur zijn technologie had gestut met zoveel zinloze dystopie, hadden diens recensenten geklaagd over ongeloofwaardi… | Als een sciencefictionauteur zijn technologie had gestut met zoveel zinloze dystopie, hadden diens recensenten geklaagd over ongeloofw… | ↗Not briefed |
hn hn-ai | 4d | Show HN: Blue-harness – audit and harden binaries against AI-assisted reversing | — | ↗Not briefed |
bluesky bluesky-ai | 4d | 3/ Run your own workload parameters & interactive FinOps stress test in our full engineering guide: https://tokenmargin.dev/guides/ai-saas-… | 3/ Run your own workload parameters & interactive FinOps stress test in our full engineering guide: https://tokenmargin.dev/guides/ai-… | ↗Not briefed |
arxiv arxiv-ai | 4d | MetaAct: Self-Evolving Executable Action Constraints for Embodied Language Agents | Self-evolving language agents have emerged as a promising paradigm for continuously improving agent behavior through experience. Howev… | ↗Not briefed |
arxiv arxiv-ai | 4d | RAMP: Rethinking Multimodal Evaluation through Answer Readout Protocols | Multimodal large language models (MLLMs) are often evaluated by converting model outputs into discrete answers. The conversion protoco… | ↗Not briefed |
bluesky bluesky-ai | 4d | What persuaded investors to put $1.2B into Mirendil just a couple of months after its debut? The case for backing it partly rests on work … | What persuaded investors to put $1.2B into Mirendil just a couple of months after its debut? The case for backing it partly rests on… | ↗Not briefed |
arxiv arxiv-ai | 4d | Instrumenting Failure: Scoring Language Model Agents at Every Decision Node in Imperfect-Information Games | Agent evaluations report task success and rarely say what an agent did at each decision. We score every decision node of three 12B to… | ↗Not briefed |
arxiv arxiv-ai | 4d | Safety-Preserving Medical LLM Fine-Tuning via Bilevel Data Selection | As general large language models (LLMs) are increasingly adapted for patient-facing medical applications, improving medical capability… | ↗Not briefed |
bluesky bluesky-ai | 4d | Mistral verfolgt einen ganz anderen Usecase als z.B. OpenAI oder Anthropic. Wer Wert auf Datensouveränität, onpremise Hosting und zugeschn… | Mistral verfolgt einen ganz anderen Usecase als z.B. OpenAI oder Anthropic. Wer Wert auf Datensouveränität, onpremise Hosting und zug… | ↗Not briefed |
hn hn-ai | 4d | Show HN: Brief2mp4 – ClaudeCode/Codex skill that checks videos against the brief | — | ↗Not briefed |
arxiv arxiv-ai | 4d | Contrastive Activation-Guided Zeroth-Order Preference Optimisation for Memory-Efficient LLM Alignment | Direct Preference Optimisation (DPO) simplifies reinforcement learning from human feedback, but still relies on traditional backpropag… | ↗Not briefed |
rss techmeme | 4d | Copenhagen-based AI accounting software maker Sumary raised a $25M seed led by Singular, bringing total funding to $29.2M after a $4.2M pre-seed in Nov. 2025 (Sofia Chesnokova/Tech Funding News) | Sofia Chesnokova / Tech Funding News: Copenhagen-based AI accounting software maker Sumary raised a $25M seed led by Singular, bringin… | ↗Not briefed |
arxiv arxiv-ai | 4d | Instructions Before Actions: Prospective and Retrospective Language Supervision for Driving Vision-Language-Action Data | Driving vision-language-action (VLA) models can accept navigation instructions, yet predicting a driver's trajectory does not establis… | ↗Not briefed |
hn hn-ai | 4d | S3rg: Grep for object storage, written in Rust | — | ↗Not briefed |
bluesky bluesky-ai | 4d | Deepseek最近限额炸出了好多API男大,一边说女性只会跟deepseek聊天,浪费算力;一边夸赞API便宜比订阅制划算。(更不用说最经典的写代码>情感陪伴) 用个破API真的会让人自豪感暴增,还成为性别优越性的象征了 | Deepseek最近限额炸出了好多API男大,一边说女性只会跟deepseek聊天,浪费算力;一边夸赞API便宜比订阅制划算。(更不用说最经典的写代码>情感陪伴) 用个破API真的会让人自豪感暴增,还成为性别优越性的象征了 | ↗Not briefed |
hn hn-ai | 4d | Capsules: Shared memory your agents can hand off – and you can read | — | ↗Not briefed |
bluesky bluesky-ai | 4d | ただでさえゴミみたいな頭してる上に返信をLLMに書かせてる疑いがあるので先行ブロックをお勧めする。 | ただでさえゴミみたいな頭してる上に返信をLLMに書かせてる疑いがあるので先行ブロックをお勧めする。 | ↗Not briefed |
arxiv arxiv-ai | 4d | GPromptShield: Elevating Resilience in Graph Prompt Tuning against Adversarial Attacks | The paradigm of ``pre-training and prompt-tuning", with its effectiveness and lightweight characteristics, has rapidly spread from the… | ↗Not briefed |
hn hn-ai | 4d | TrustGraph – ACLs that follow the data through your AI agents | — | ↗Not briefed |
arxiv arxiv-ai | 4d | GEAR$^2$: Pruned Experts Live On through Gated Reuse of Activated Experts and a Missing-Mass Residual | Mixture-of-Experts (MoE) Large Language Models (LLMs) activate only a few experts per token while generally keeping all the experts in… | ↗Not briefed |
hn hn-ai | 4d | Hibana: Fast Agentic Inference on Unified-Memory Devices | — | ↗Not briefed |
bluesky bluesky-ai | 4d | now show me deepseek | now show me deepseek | ↗Not briefed |
bluesky bluesky-ai | 4d | Stocks extended losses overnight as traders await Nvidia earnings this week and Anthropic IPO developments. | Stocks extended losses overnight as traders await Nvidia earnings this week and Anthropic IPO developments. | ↗Not briefed |
rss techmeme | 4d | Source: Musk will attend this week's APEC summit in Shenzhen, where Trump, Xi, and Putin will meet; many US business leaders may skip over scheduling conflicts (Nikkei Asia) | Nikkei Asia: Source: Musk will attend this week's APEC summit in Shenzhen, where Trump, Xi, and Putin will meet; many US business lead… | ↗Not briefed |
arxiv arxiv-ai | 4d | Newton-PINet: A fast physics-informed neural network with Newton linearization for meta-learning nonlinear PDEs | Scientific machine learning has opened new avenues for solving parameterized partial differential equations (PDEs), enabling models to… | ↗Not briefed |
hn hn-ai | 4d | cudamemcpy-tuner: cudaMemcpying up to 5x faster on a B300 | — | ↗Not briefed |