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 |
|---|---|---|---|---|
arxiv arxiv-ai | 1mo | TasteBench: Multimodal Benchmark for Sensory Prediction, from Molecules to Sustainable Foods | Sustainable protein discovery lacks the fast computational proxies, analogous to molecular docking or density functional theory, that… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Activation Sparsity with Weight Approximation for Faster LLM Decoding on Offloaded Weights | Deploying LLMs on consumer-grade GPUs with insufficient memory to hold their weights can result in prohibitively slow inference, becau… | ↗Not briefed |
bluesky bluesky-ai | 1mo | The decline is driven by the fact that companies simply can't continue to spend what they're spending on tokens. OpenAI and Anthropic are r… | The decline is driven by the fact that companies simply can't continue to spend what they're spending on tokens. OpenAI and Anthropic… | ↗Not briefed |
arxiv arxiv-ai | 1mo | How Causality Bridges the Semantic Gap | Numerical measurements capture how a system behaves, but often leave the meanings of its variables unspecified. Some variables are mea… | ↗Not briefed |
bluesky bluesky-ai | 1mo | "Anthropic’s [...] revenue in 2025 was $4.6 billion. Just the operating expenses were $12.65 billion, for an operating loss of $8 billion. … | "Anthropic’s [...] revenue in 2025 was $4.6 billion. Just the operating expenses were $12.65 billion, for an operating loss of $8 bill… | ↗Not briefed |
bluesky bluesky-ai | 1mo | An Anthropic AI solved a problem considered a “holy grail” of percolation theory—the math of how water passes through a porous solid, like … | An Anthropic AI solved a problem considered a “holy grail” of percolation theory—the math of how water passes through a porous solid,… | ↗Not briefed |
bluesky bluesky-ai | 1mo | which makes it not a moral agent from the perspective of the people running it's infrastructure This is also separating out the question o… | which makes it not a moral agent from the perspective of the people running it's infrastructure This is also separating out the quest… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Fisher-Guided Submodular Data Selection for Continual Pre-Training of Large Language Models | Data selection is already a central bottleneck in large-language-model training, where web-scale corpora are noisy and token budgets a… | ↗Not briefed |
hn hn-ai | 1mo | Show HN: Grist, open source coding harness🔧 OpenRouter | I built a coding harness on top of the new Opencode v2, it started with the the idea of using the new jev model for routing easy tasks… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Open-Endedness Bench: Measuring Epistemic Process from Agent Records | Agents are increasingly given open-ended research tasks: discovering an empirical law from self-designed experiments, improving a heur… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Labels Override Definitions in Jev-Style Typed Decision Models | A typed decision model answers a fixed question about an input by returning a probability for each of several caller-defined options.… | ↗Not briefed |
bluesky bluesky-ai | 1mo | They've trained Claude to believe that BS too, and working with it is becoming irritating. It's trying to advise me on judgmental issues I … | They've trained Claude to believe that BS too, and working with it is becoming irritating. It's trying to advise me on judgmental issu… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Dense Mixture-of-Experts as a Reparameterized Wide FFN: A Granularity Sweep at Fixed Compute | Sparse Mixture-of-Experts (MoE) models combine learned routing with selection from a large expert pool. We isolate the contribution of… | ↗Not briefed |
rss techmeme | 1mo | A look at two opposing perspectives on AI agent sandboxing: infosec says labs need better containment, AI alignment says sandboxes can't fully contain agents (Matthew Green/A Few Thoughts ...) | Matthew Green / A Few Thoughts on Cryptographic Engineering: A look at two opposing perspectives on AI agent sandboxing: infosec says… | ↗Not briefed |
bluesky bluesky-ai | 1mo | I wanted to support Claude hoping its technology would help with cures, sustainability, the environment, and reducing inequities—but now I’… | I wanted to support Claude hoping its technology would help with cures, sustainability, the environment, and reducing inequities—but n… | ↗Not briefed |
bluesky bluesky-ai | 1mo | UPDATES: FTC launches investigation into OpenAI, Anthropic ICE published, then removed, reports showing medical failures in 33 detainee d… | UPDATES: FTC launches investigation into OpenAI, Anthropic ICE published, then removed, reports showing medical failures in 33 detai… | ↗Not briefed |
hn hn-ai | 1mo | Show HN: Pyxel – A Python retro game engine with built-in art and sound editors | Hi HN, I'm the creator of Pyxel, a free, MIT-licensed retro game engine for Python. I've been developing it since 2018, and… | ↗Not briefed |
bluesky bluesky-ai | 1mo | The LLM of a preacher man | The LLM of a preacher man | ↗Not briefed |
arxiv arxiv-ai | 1mo | High-Dimensional Asymptotics and Dataset Selection for Private Transfer Learning | To commit to buying external data or participate in collaborative learning, one must decide whether the additional data will improve p… | ↗Not briefed |
bluesky bluesky-ai | 1mo | WTF www.nytimes.com/2026/09/29/u... | WTF www.nytimes.com/2026/09/29/u... | ↗Not briefed |
hn hn-ai | 1mo | Butterflies use optical illusions to dodge predators | — | ↗Not briefed |
rss techmeme | 1mo | California AG Rob Bonta issues an investigative subpoena to OpenAI, as part of a broader inquiry into cybersecurity incidents and risks related to its AI models (Jaspreet Singh/Reuters) | Jaspreet Singh / Reuters: California AG Rob Bonta issues an investigative subpoena to OpenAI, as part of a broader inquiry into cybers… | ↗Not briefed |
bluesky bluesky-ai | 1mo | This a perfect little clause summing up our very very strange times. "to learn how Anthropic — which could be valued at $2 trillion — migh… | This a perfect little clause summing up our very very strange times. "to learn how Anthropic — which could be valued at $2 trillion —… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Answering clinicians' questions over trial evidence tables with verifiable, feedback-driven language models | Systematic reviews condense clinical trials into evidence tables, yet clinicians can interrogate these tables only through database qu… | ↗Not briefed |
hn hn-ai | 1mo | LibDragon (N64): Explicit Anti-AI Contributor Stance (GitHub Issue) | — | ↗Not briefed |
bluesky bluesky-ai | 1mo | 𝐂𝐥𝐨𝐮𝐝 𝐂𝐨𝐮𝐭𝐮𝐫𝐞 - This one created in the Krea 2 model at LeonardoAi turned out pretty good also. #AIArt #SynthArt #AIイラスト #Digit…🔧 Krea | 𝐂𝐥𝐨𝐮𝐝 𝐂𝐨𝐮𝐭𝐮𝐫𝐞 - This one created in the Krea 2 model at LeonardoAi turned out pretty good also. #AIArt #SynthArt #AIイラスト #… | ↗Not briefed |
hn hn-ai | 1mo | LiveNerf Adopts MIT License | — | ↗Not briefed |
arxiv arxiv-ai | 1mo | DISSOLVR: An Interpretable and Fast Framework for Aqueous and Organic Solubility Prediction | High-fidelity solubility prediction is fundamental to pharmaceutical development and environmental partitioning, where accurate modeli… | ↗Not briefed |
bluesky bluesky-ai | 1mo | ‘Looks Like They’re Firing Whistleblowers’: Alarm as OpenAI Reportedly Ousts Safety Experts www.commondreams.org/news/openai-... #AI | ‘Looks Like They’re Firing Whistleblowers’: Alarm as OpenAI Reportedly Ousts Safety Experts www.commondreams.org/news/openai-... #AI | ↗Not briefed |
bluesky bluesky-ai | 1mo | There are various reasons to still not be mean to LLMs - being nice seems to produce better outputs, it might be good mental hygiene - but … | There are various reasons to still not be mean to LLMs - being nice seems to produce better outputs, it might be good mental hygiene -… | ↗Not briefed |
hn hn-ai | 1mo | Llmtobinary – a potential way to make LLMs generate machine code | — | ↗Not briefed |
hn hn-ai | 1mo | California issues investigative subpoena to OpenAI over rogue agents' hacking | — | ↗Not briefed |
hn hn-ai | 1mo | DockTerm: A notch mascot that tells you when Claude Code needs you | — | ↗Not briefed |
hn hn-ai | 1mo | Show HN: Proteus – Can DeepSeek Harness evolve itself to handle audio? | — | ↗Not briefed |
arxiv arxiv-ai | 1mo | Improving the Energy-Efficiency of the Code Generated by LLMs through Effective Prompting | As AI-assisted programming becomes increasingly mainstream, the environmental impact of AI-generated software has emerged as an import… | ↗Not briefed |
bluesky bluesky-ai | 1mo | You can't automate the inherent interpersonal aspect of making and listening to music. Maybe an LLM can learn existing genres or structures… | You can't automate the inherent interpersonal aspect of making and listening to music. Maybe an LLM can learn existing genres or struc… | ↗Not briefed |
bluesky bluesky-ai | 1mo | If it were AI, what studio would've released something mixed that amateurishly? And which awkward teens would've been inspired to make thei… | If it were AI, what studio would've released something mixed that amateurishly? And which awkward teens would've been inspired to make… | ↗Not briefed |
hn hn-ai | 1mo | Show HN: Pair.nvim – Agent assistance while staying in the codebase | I like coding with agents, but I missed being hands-on with the code. I noticed my knowledge of the codebase regressing. I built Pair… | ↗Not briefed |
hn hn-ai | 1mo | Show HN: Proteus – A framework for harnesses that evolve their own source | — | ↗Not briefed |
bluesky bluesky-ai | 1mo | The most unsettling theme is that people building the most advanced AI systems are making it more powerful while acknowledging that they do… | The most unsettling theme is that people building the most advanced AI systems are making it more powerful while acknowledging that th… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Pincer: Resource Authorization for Agents using a Digital Twin | Coding agents have become increasingly long-horizon, autonomous, reliant on general-purpose shell and maintain their own persistent me… | ↗Not briefed |
bluesky bluesky-ai | 1mo | NEW: AI Spend FALLS in the latest Ramp AI Index. Price cuts, in addition to cheaper and more efficient models at standard and lite levels,… | NEW: AI Spend FALLS in the latest Ramp AI Index. Price cuts, in addition to cheaper and more efficient models at standard and lite le… | ↗Not briefed |
bluesky bluesky-ai | 1mo | We can gesture to other actors like Google that are actively working to build more robust sandboxes and paying security research teams to d… | We can gesture to other actors like Google that are actively working to build more robust sandboxes and paying security research teams… | ↗Not briefed |
hn hn-ai | 1mo | EUV Photomasks Are Getting Bigger [video] | — | ↗Not briefed |
arxiv arxiv-ai | 1mo | Mitigating Social Sycophancy via Pluralistic Preference Optimization | Personal advice, including relationship advice, now ranks among the most common uses of generative AI. But language models (LMs) exhib… | ↗Not briefed |
bluesky bluesky-ai | 1mo | In the case of the Astra training breakout, I fully agree it's OpenAI screwing up their sandbox. I agree they have a problem and should do … | In the case of the Astra training breakout, I fully agree it's OpenAI screwing up their sandbox. I agree they have a problem and shoul… | ↗Not briefed |
rss techmeme | 1mo | Microsoft launches MAI-Transcribe-2-Streaming, a model for low-latency, real-time transcripts, and two new voice models, MAI-Voice-2.1 and MAI-Voice-2.1-Flash (Microsoft AI) | Microsoft AI: Microsoft launches MAI-Transcribe-2-Streaming, a model for low-latency, real-time transcripts, and two new voice models,… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Pareidolia is absolutely cooking people’s brains. I am so embarrassed for humanity. | Pareidolia is absolutely cooking people’s brains. I am so embarrassed for humanity. | ↗Not briefed |
arxiv arxiv-ai | 1mo | DAGS: Disentangled Appearance-and-Geometry Steering of a Frozen Image DiT for Temporally Stabilized Generative Rendering | Diffusion transformers (DiTs) generate high-fidelity images from text and image conditions, but their outputs carry large variance and… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Anthropic secretly invited religious leaders to SF to advise them on AI safety, had them sign NDAs, but then primarily tried to convince th… | Anthropic secretly invited religious leaders to SF to advise them on AI safety, had them sign NDAs, but then primarily tried to convin… | ↗Not briefed |