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 | 26d | Show HN: A curated directory of AI Tools | A curated directory of AI-powered coding tools, autonomous agents, and developer utilities. | ↗Not briefed |
rss techmeme | 26d | Alexandr Wang says ex-Apple exec Daniel Gross, now at Meta, helped bring Meta and Apple together: "we're excited for mommy and daddy to be playing nice again" (Ina Fried/Axios) | Ina Fried / Axios: Alexandr Wang says ex-Apple exec Daniel Gross, now at Meta, helped bring Meta and Apple together: “we're excited fo… | ↗Not briefed |
arxiv arxiv-ai | 26d | BitSparse: Compact Activation Storage for ReLU Feed-Forward Networks | Reducing memory use in large language model (LLM) training is an ongoing challenge. Activation checkpointing saves memory by recomputa… | ↗Not briefed |
bluesky bluesky-ai | 26d | You’re free to find work elsewhere. One preferably away from a computer or phone so you’re not tempted to use an LLM to ask if it’s feelin… | You’re free to find work elsewhere. One preferably away from a computer or phone so you’re not tempted to use an LLM to ask if it’s f… | ↗Not briefed |
arxiv arxiv-ai | 26d | Explore Once, Evaluate Many: Multiple Policy Evaluation for Stochastic Linear Bandits | We study how to evaluate $K$ target policies simultaneously in stochastic contextual bandits with linear reward structure. Given a kno… | ↗Not briefed |
arxiv arxiv-ai | 26d | Not X, but When: Diagnosing a Mandarin LLM Writing Tic through Contextual Appropriateness | Recurring expressions in model-written text are often treated as signatures of machine authorship, yet most of them serve legitimate f… | ↗Not briefed |
bluesky bluesky-ai | 26d | The thought that someone will likely *seriously* suggest gender-affirming care for LLM bots before humans sure is something. | The thought that someone will likely *seriously* suggest gender-affirming care for LLM bots before humans sure is something. | ↗Not briefed |
rss techmeme | 26d | Meta says it worked with Apple to make Muse Spark Open run locally on M5 Ultra Mac Studios, letting users pair two to four Studios to code or do agentic work (CNBC) | CNBC: Meta says it worked with Apple to make Muse Spark Open run locally on M5 Ultra Mac Studios, letting users pair two to four Studi… | ↗Not briefed |
arxiv arxiv-ai | 26d | Disagreement-Guided Weak-to-Strong Routing | Small language models are increasingly deployed on edge devices and robotic systems to meet cost and latency constraints, yet their li… | ↗Not briefed |
arxiv arxiv-ai | 26d | Language Is a Security Surface for Agent Guards | Step-level guards decide, before each tool call, whether an agent's next action is safe given the user's request and the trajectory so… | ↗Not briefed |
bluesky bluesky-ai | 26d | very adorable that people on here think valve would never use AI when gabe newell was donating to openai in 2018 and was saying last year t… | very adorable that people on here think valve would never use AI when gabe newell was donating to openai in 2018 and was saying last y… | ↗Not briefed |
bluesky bluesky-ai | 26d | ‘Every investment boom brings with it a rogues’ gallery of acronyms and valuation techniques, sometimes designed to elucidate and other tim… | ‘Every investment boom brings with it a rogues’ gallery of acronyms and valuation techniques, sometimes designed to elucidate and othe… | ↗Not briefed |
arxiv arxiv-ai | 26d | Shortcut Sensitivity in Agent-Safety Evaluations | AI agents take actions through tools. Some actions are sensitive, so guards that supervise them must themselves be evaluated reliably.… | ↗Not briefed |
bluesky bluesky-ai | 26d | Porra Metrópolis anti LLM fazendo edição com humanos isso ai | Porra Metrópolis anti LLM fazendo edição com humanos isso ai | ↗Not briefed |
arxiv arxiv-ai | 26d | SurvICL: a Foundation Model for in-context Survival Analysis | Tabular foundation models (TFMs) now outperform gradient-boosted trees in classification and regression. Yet in survival analysis, whi… | ↗Not briefed |
bluesky bluesky-ai | 26d | the reason something like an LLM is interesting from a phenomenological perspective is precisely the same reason that a book like solaris i… | the reason something like an LLM is interesting from a phenomenological perspective is precisely the same reason that a book like sola… | ↗Not briefed |
bluesky bluesky-ai | 26d | Not exact but this on AI failures gets at same idea. It's not a "failure" except that we expect the model to do something that it precludes… | Not exact but this on AI failures gets at same idea. It's not a "failure" except that we expect the model to do something that it prec… | ↗Not briefed |
hn hn-ai | 26d | Show HN: Sieve - a small homelab scanner for CVES, root passwords and ESP boards | — | ↗Not briefed |
arxiv arxiv-ai | 26d | Query-Efficient Rejection Sampling for Parallel Masked Diffusion | Masked diffusion models (MDMs) accelerate generation by unmasking multiple tokens in parallel, but sampling these tokens independently… | ↗Not briefed |
arxiv arxiv-ai | 26d | Training-Free On-Policy Distillation | On-policy distillation consolidates specialist models into one student at the price of rollouts, teacher passes and gradient steps for… | ↗Not briefed |
arxiv arxiv-ai | 26d | OS-Helix: Unifying GUI Interaction and Code Execution in Native Computer-Using Agents | Real desktop automation is neither purely visual nor purely programmatic. Agents must understand user-facing interfaces while efficien… | ↗Not briefed |
hn hn-ai | 26d | Local and Distributed Large LLM Inference with MLX and RDMA on Mac Studio | — | ↗Not briefed |
rss techmeme | 26d | Meta releases Muse Spark Open, saying it is the most capable US open-weight model, with 816B total and 42B active parameters and a 1M-token context window (Meta AI Research) | Meta AI Research: Meta releases Muse Spark Open, saying it is the most capable US open-weight model, with 816B total and 42B active pa… | ↗Not briefed |
arxiv arxiv-ai | 26d | Architecting for Hardware Diversity: Meta's End-to-End Co-Design for High-Performance AI Training on AMD GPUs | Meta deploys a heterogeneous fleet comprising NVIDIA GPUs, AMD GPUs, and proprietary MTIA accelerators to reduce costs and mitigate ve… | ↗Not briefed |
hn hn-ai | 26d | A Claude Code skill that takes an iOS app from code to the App Store | — | ↗Not briefed |
bluesky bluesky-ai | 26d | AI goth model with 15 million OF followers starts the "Cut Off Your Leg And Mail It To Me Challenge", resulting in significant US military … | AI goth model with 15 million OF followers starts the "Cut Off Your Leg And Mail It To Me Challenge", resulting in significant US mili… | ↗Not briefed |
hn hn-ai | 26d | Photopea DMCA Takedown on GitHub | — | ↗Not briefed |
hn hn-ai | 26d | Show HN: VX – Vite-like task runner. Fastest, safest, open, yours. | — | ↗Not briefed |
bluesky bluesky-ai | 26d | Eight states are suing OpenAI. | Eight states are suing OpenAI. | ↗Not briefed |
hn hn-ai | 26d | Team of 5 post-trains GLM 5.3 Flash to Fable 5 class for $300k all-in | — | ↗Not briefed |
hn hn-ai | 26d | More on Anthropic's Scion ripping off Mononoke | — | ↗Not briefed |
bluesky bluesky-ai | 26d | Update on yesterday's post about Anthropic ripping off Mononoke: | Update on yesterday's post about Anthropic ripping off Mononoke: | ↗Not briefed |
bluesky bluesky-ai | 26d | you can make full-precision deepseek v4.1 flash fit as well (and at similar speeds as the v4!) if you’re brave enough | you can make full-precision deepseek v4.1 flash fit as well (and at similar speeds as the v4!) if you’re brave enough | ↗Not briefed |
bluesky bluesky-ai | 26d | LLM-era comment spam makes me sad. Comment spam used to be mostly super-generic. If you had any familiarity with it at all, you could ide… | LLM-era comment spam makes me sad. Comment spam used to be mostly super-generic. If you had any familiarity with it at all, you coul… | ↗Not briefed |
hn hn-ai | 26d | Whathappened.here – a social check-in app built on top of OpenStreetMap | — | ↗Not briefed |
arxiv arxiv-ai | 26d | How Much Physics Can Vision Learn from Audio? Probing and Distilling Damping Structure | Physical understanding is crucial for decision making, yet vision models struggle to infer properties such as damping that govern how… | ↗Not briefed |
bluesky bluesky-ai | 26d | anyway this is why i think anthropic banning people for being abusive to the chatbots is a good thing. burning tokens on this is just not a… | anyway this is why i think anthropic banning people for being abusive to the chatbots is a good thing. burning tokens on this is just… | ↗Not briefed |
bluesky bluesky-ai | 26d | you can make deepseek v4 flash fit iirc, as well as some larger qwens | you can make deepseek v4 flash fit iirc, as well as some larger qwens | ↗Not briefed |
bluesky bluesky-ai | 26d | teach it, sure. which i say somewhere in this thread. but it cannot be used *on* gen ai and that's intentional. literacy frameworks cannot … | teach it, sure. which i say somewhere in this thread. but it cannot be used *on* gen ai and that's intentional. literacy frameworks ca… | ↗Not briefed |
arxiv arxiv-ai | 26d | Skill-Decoupled On-Policy Expert Distillation for Unified Vision-Language Models🔧 Elicit | Reinforcement learning (RL) can elicit specialized visual capabilities in vision-language models (VLMs), yet jointly optimizing hetero… | ↗Not briefed |
arxiv arxiv-ai | 26d | FutureLift: Lifting Trajectory Proposals into Bidirectional Future Graphs for Motion Forecasting | Accurate motion prediction is critical for safe autonomous driving, yet even strong forecasters leave residual errors in their multimo… | ↗Not briefed |
bluesky bluesky-ai | 26d | The attorney for the former OpenAI employee accused of planting a recording device in the company’s San Francisco headquarters is calling t… | The attorney for the former OpenAI employee accused of planting a recording device in the company’s San Francisco headquarters is call… | ↗Not briefed |
arxiv arxiv-ai | 26d | GNN-Guided Local Refinement for Efficient Layout Decomposition with DSA-MP | Multiple patterning lithography (MPL) has been widely used to improve lithography resolution in advanced nodes, but it faces cost limi… | ↗Not briefed |
arxiv arxiv-ai | 26d | Generalization Guarantees of Randomized Block Coordinate Descent for Nonconvex Learning | Randomized block coordinate descent (RBCD) is a popular optimization algorithm for many large-scale machine learning problems. The exi… | ↗Not briefed |
arxiv arxiv-ai | 26d | TRACE: Topology-Aware Reconnection via Prediction-Adaptive Connectivity Enforcement for Thin-Structure Segmentation | Thin, elongated structures are particularly sensitive to connectivity errors, as even a few local segmentation mistakes can disrupt ot… | ↗Not briefed |
hn hn-ai | 26d | List of email domains used by AI agents | — | ↗Not briefed |
arxiv arxiv-ai | 26d | Beyond Downstream QA: Rethinking How We Evaluate Agent Memory | Agent memory benchmarks evaluate whether an agent's responses reflect correct knowledge, measuring downstream question-answering accur… | ↗Not briefed |
bluesky bluesky-ai | 26d | media literacy depends on ontologies like sources, process, theories, taxonomies -- all things gen ai obscures in its corporate ownership o… | media literacy depends on ontologies like sources, process, theories, taxonomies -- all things gen ai obscures in its corporate owners… | ↗Not briefed |
rss the-decoder | 26d | Black Forest Labs' open-weights image model FLUX 3 Image Klein runs on consumer GPUs with just 8 GB of VRAM🔧 FLUX | Black Forest Labs has released FLUX 3 Image Klein, a 5.6-billion-parameter open-weights image model that supports native 4K generation… | ↗Not briefed |
bluesky bluesky-ai | 26d | Still following the promise I made to myself when I joined Bluesky: block dumb-asses instead of commenting or stressing on them. LLM consci… | Still following the promise I made to myself when I joined Bluesky: block dumb-asses instead of commenting or stressing on them. LLM c… | ↗Not briefed |