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 | 1mo | LocalThings: SmartThings devices in HomeAssistant without a cloud connection | — | ↗Not briefed |
arxiv arxiv-ai | 1mo | Beyond Accuracy: Prefix-Invariant Realizations of Low-Precision Fast Matrix Multiplication | Fast matrix multiplication saves multiplications through exact cancellation, but rounding sums that mix token rows can leave contribut… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Given enough training data, a frontier AI model could likely produce EDM indistinguishable from someone like David Guetta. Yet no AI model … | Given enough training data, a frontier AI model could likely produce EDM indistinguishable from someone like David Guetta. Yet no AI m… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Backward-State Policy Is Part of the Learning Algorithm | Low-precision training rounds tensors that the backward pass reads again, often for several gradients; each use can read the forward's… | ↗Not briefed |
hn hn-ai | 1mo | Sub-two-second LLM-driven NPC behavior in a 3D action game | — | ↗Not briefed |
arxiv arxiv-ai | 1mo | A Comprehensive Benchmark of Source-Free Universal Domain Adaptation on Time Series Representations | Source-Free Universal Domain Adaptation (SF-UniDA) extends Universal Domain Adaptation by removing access to source data at adaptation… | ↗Not briefed |
bluesky bluesky-ai | 1mo | i’d be curious what OpenAI’s internal benchmarks test for. And Anthropic’s my hunch is both do better on their own benchmarks | i’d be curious what OpenAI’s internal benchmarks test for. And Anthropic’s my hunch is both do better on their own benchmarks | ↗Not briefed |
arxiv arxiv-ai | 1mo | Stress-Testing LLM Lie Detectors: Role-Play Failures and Spurious Correlations | Lie detection probes aim to predict from a language model's internal states whether its output is truthful or dishonest. However, role… | ↗Not briefed |
bluesky bluesky-ai | 1mo | It matters for theory because if LLM can do advanced math without tools, that disproves the claim that it's just shuffling tokens around. … | It matters for theory because if LLM can do advanced math without tools, that disproves the claim that it's just shuffling tokens arou… | ↗Not briefed |
youtube youtube-mattvidpro | 1mo | I Didn't Expect YuE2 to Sound THIS Good🔧 Suno | Hello all you fantastic viewers and welcome back to the Matt Vid Pro channel. Today's video has me pretty hyped. We're comparing Suno… | ↗Not briefed |
bluesky bluesky-ai | 1mo | teaching a unit on AI interpretability and something I notice doesn't seem to be discussed much: wouldn't figuring out interpretability rai… | teaching a unit on AI interpretability and something I notice doesn't seem to be discussed much: wouldn't figuring out interpretabilit… | ↗Not briefed |
hn hn-ai | 1mo | DreamZero: World Action Models Are Zero-Shot Policies | — | ↗Not briefed |
bluesky bluesky-ai | 1mo | FTC opens probe into AI giants including Anthropic and OpenAI, source says reut.rs/4hxKOah | FTC opens probe into AI giants including Anthropic and OpenAI, source says reut.rs/4hxKOah | ↗Not briefed |
rss techmeme | 1mo | Sources: the US FTC is expanding a sweeping probe of Anthropic, OpenAI, and other frontier AI labs, and plans to issue formal demands to turn over information (New York Post) | New York Post: Sources: the US FTC is expanding a sweeping probe of Anthropic, OpenAI, and other frontier AI labs, and plans to issue… | ↗Not briefed |
rss techcrunch-ai | 1mo | 3 days left to exhibit: Turn visibility into your next opportunity at TechCrunch Disrupt 2026 | Three days left to exhibit at TechCrunch Disrupt 2026. Book by October 2 at 11:59 p.m. PT to showcase your startup to 10,000+ founders… | ↗Not briefed |
arxiv arxiv-ai | 1mo | RATIO: Reasoning Analysis and Token-level Inference Optimization for Quantized Reasoning Models | Post-training quantization (PTQ) has become a widely adopted technique for reducing the memory footprint and inference cost of large l… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Finite-Horizon Fisher Memory in Two-Sided Power-Bounded Recurrent Systems | We analyse allocation, admission and post-write retention in finite-horizon linear-Gaussian noisy recurrent memories. At every horizon… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Probabilistic Adversarial Training | Building on a probabilistic perspective in which adversarial examples arise from the overlap between a distance-based distribution $p_… | ↗Not briefed |
bluesky bluesky-ai | 1mo | If the LLM were to solve math problems "in its head" that would be a refutation of what Colin calls the "strong stochastic parrot thesis", … | If the LLM were to solve math problems "in its head" that would be a refutation of what Colin calls the "strong stochastic parrot thes… | ↗Not briefed |
arxiv arxiv-ai | 1mo | TopTimeNet: Topologically-assisted time-series classification model | Distinguishing periodic from chaotic dynamics in a time series is a fundamental challenge in both physics and engineering. Yet, end-to… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Vous vous rappelez quand on se moquait des vidéos IA de Will Smith ? Bha en fait fin 2026 on en est là ⤵️ C’est flippant la vitesse de pro… | Vous vous rappelez quand on se moquait des vidéos IA de Will Smith ? Bha en fait fin 2026 on en est là ⤵️ C’est flippant la vitesse d… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Pseudo-Label-Triggered Retraining from Forecast Errors for Online Time Series Forecasting | Real-world time series forecasting systems operate under non-stationary data streams, where forecasting performance may degrade over t… | ↗Not briefed |
rss techmeme | 1mo | Amazon aims to have 5,000 drivers wearing its smart glasses in 2026, rising to 20,000+ in 2027; the glasses capture static images for its AI mapping platform (Dave Lee/Bloomberg) | Dave Lee / Bloomberg: Amazon aims to have 5,000 drivers wearing its smart glasses in 2026, rising to 20,000+ in 2027; the glasses capt… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Safety of Latent Communication in Multi-Agent Systems | Latent communication enables multi-agent systems to exchange information directly in internal representation space, reducing the token… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Explore-on-Graph: Hybrid Embedding-LLM Reasoning for Knowledge Graph Question Answering under Incompleteness | Large language models (LLMs) are increasingly combined with knowledge graphs (KGs) to ground reasoning in structured evidence. However… | ↗Not briefed |
bluesky bluesky-ai | 1mo | I don't know if there's a solid legal claim here and I'm certainly skeptical of Uthmeier, but this is a helluva lede: "[OpenAI/Sam Altman]… | I don't know if there's a solid legal claim here and I'm certainly skeptical of Uthmeier, but this is a helluva lede: "[OpenAI/Sam Al… | ↗Not briefed |
arxiv arxiv-ai | 1mo | CORD: Learning Reusable Degradation Representations Across Heterogeneous Physical Systems | Can heterogeneous physical degradation systems benefit from joint pretraining and move beyond system-specific prognostics toward reusa… | ↗Not briefed |
bluesky bluesky-ai | 1mo | this makes it sound like the ai agent acceded something at data.cityofchicago.org. seems weird that would trigger a disclosure from openai | this makes it sound like the ai agent acceded something at data.cityofchicago.org. seems weird that would trigger a disclosure from op… | ↗Not briefed |
hn hn-ai | 1mo | Every list on GitHub, searchable, scored and kept fresh | — | ↗Not briefed |
bluesky bluesky-ai | 1mo | Like...it's not making music. That's not what music is. An LLM could write you a paragraph describing an orgasm, but it's not having one. Y… | Like...it's not making music. That's not what music is. An LLM could write you a paragraph describing an orgasm, but it's not having o… | ↗Not briefed |
arxiv arxiv-ai | 1mo | GraphMAS: A Systematic Benchmark of Multi-Agent Coordination for Graph Learning | LLM-based multi-agent systems coordinate specialized reasoning through aggregation, interaction, and adaptive control, yet their poten… | ↗Not briefed |
rss bing-news-ai | 1mo | ‘Super Intelligence,’ Trump’s Dopey AI Rebrand, Explained | Americans’ concerns about artificial intelligence have quickly escalated from “it might take our jobs” to “it might kill us all.” In S… | ↗Not briefed |
bluesky bluesky-ai | 1mo | i don’t totally understand what this story is saying. an ai agent accessed the city’s public data portal? or something else? i don’t see… | i don’t totally understand what this story is saying. an ai agent accessed the city’s public data portal? or something else? i don’… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Riemannian Flow Models with Reinforcement Learning for Molecular Crystal Structure Prediction | Crystal structure governs material properties, making crystal structure prediction (CSP) a fundamental problem in materials science. G… | ↗Not briefed |
bluesky bluesky-ai | 1mo | I don't really understand the screenshotted thread as a critique of LLMs. If you ask an LLM a math problem easily solved by a calculator it… | I don't really understand the screenshotted thread as a critique of LLMs. If you ask an LLM a math problem easily solved by a calculat… | ↗Not briefed |
hn hn-ai | 1mo | Artist "interrupts sightline" of Gateway Arch with animal-hide sculpture | — | ↗Not briefed |
arxiv arxiv-ai | 1mo | Security Properties of Neural Networks as Decision Problems | Certifying a deployed neural network raises decision problems that the verification literature has not classified: whether the model c… | ↗Not briefed |
hn hn-ai | 1mo | Crap: A Fork of Poop | — | ↗Not briefed |
bluesky bluesky-ai | 1mo | This a million times! Why would I eat radioactive cat vomit while destroying democracy and the planet? My own LLM-hating has no limits. It… | This a million times! Why would I eat radioactive cat vomit while destroying democracy and the planet? My own LLM-hating has no limit… | ↗Not briefed |
arxiv arxiv-ai | 1mo | How Does Local Landscape Geometry Evolve in Language Model Pre-Training? | The scale and expense of pre-training language models make efficient hyperparameter tuning essential, yet a principled guidance is sti… | ↗Not briefed |
hn hn-ai | 1mo | Crusader Kings 3's By God Alone expansion offers plenty of holy hijinks | — | ↗Not briefed |
bluesky bluesky-ai | 1mo | i once saw the former arguing “no, that’s not an LLM. it’s a large language model, which is different” and i think i lost a few brain cells… | i once saw the former arguing “no, that’s not an LLM. it’s a large language model, which is different” and i think i lost a few brain… | ↗Not briefed |
arxiv arxiv-ai | 1mo | MemCodex: Self-Programming Hierarchical Memory for Language Agents | Agent memory faces heterogeneous access needs: a single-hop question may require one piece of evidence, whereas a multi-hop question m… | ↗Not briefed |
arxiv arxiv-ai | 1mo | DiffWAM: A Fast and Efficient Navigation World Action Model | Pretrained video foundation models encode rich semantic and spatiotemporal priors for embodied navigation, yet converting these priors… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Also *extremely* funny that the executive order renaming AI to Super Intelligence is signed by the CEOs of “OpenAI” and “xAI”. Bold move … | Also *extremely* funny that the executive order renaming AI to Super Intelligence is signed by the CEOs of “OpenAI” and “xAI”. Bold… | ↗Not briefed |
bluesky bluesky-ai | 1mo | $200 of deepseek API, assuming all your usage is offpeak, gets you.. about the 120M output tokens/115 bibles number. so: $200 deepseek ~= … | $200 of deepseek API, assuming all your usage is offpeak, gets you.. about the 120M output tokens/115 bibles number. so: $200 deepsee… | ↗Not briefed |
bluesky bluesky-ai | 1mo | (this isn't necessarily unreasonable. LLMs are kind of a weird new thing and while I think "it's a digital friend!" is actually a really ba… | (this isn't necessarily unreasonable. LLMs are kind of a weird new thing and while I think "it's a digital friend!" is actually a real… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Revisiting On-policy Adversarial Black-Box Distillation: Calibrating Groupwise Reward Geometry for Effective Advantage Construction | Black-box distillation is a practical route for transferring capabilities from API-accessible large language models that expose only t… | ↗Not briefed |
bluesky bluesky-ai | 1mo | If you're at #COLM2026, come check out our poster for "Spoiler Alert" (arxiv.org/abs/2604.09854). We ask why LLM fiction is so bad at holdi… | If you're at #COLM2026, come check out our poster for "Spoiler Alert" (arxiv.org/abs/2604.09854). We ask why LLM fiction is so bad at… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Door Anthropics AI 'ontdekte' enzymen bleken al jaren bekend bij onderzoekers die hun resultaten gedeeld hadden met... de AI van Anthropic.… | Door Anthropics AI 'ontdekte' enzymen bleken al jaren bekend bij onderzoekers die hun resultaten gedeeld hadden met... de AI van Anthr… | ↗Not briefed |