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 | 28d | SwarmRSI: Recursive Self-Improvement through Evolving Scientific Agent Swarm | Recursive self-improvement (RSI) in scientific research aims to improve agents' research capabilities through their own experimental e… | ↗Not briefed |
bluesky bluesky-ai | 28d | doesn’t Tailscale have smth like this? I think it’s what their aperture is for aperture.tailscale.com @apenwarr.ca ? | doesn’t Tailscale have smth like this? I think it’s what their aperture is for aperture.tailscale.com @apenwarr.ca ? | ↗Not briefed |
bluesky bluesky-ai | 28d | i am a little worried that open models will start actually falling behind with how much work ant (and presumably openai?) is putting into p… | i am a little worried that open models will start actually falling behind with how much work ant (and presumably openai?) is putting i… | ↗Not briefed |
bluesky bluesky-ai | 28d | I think it means it's time for new DeepSeek and GLM models :p | I think it means it's time for new DeepSeek and GLM models :p | ↗Not briefed |
bluesky bluesky-ai | 28d | I can't fucking believe this is a real thing and not a joke like. program attached to my login credentials for however many digital purchas… | I can't fucking believe this is a real thing and not a joke like. program attached to my login credentials for however many digital pu… | ↗Not briefed |
bluesky bluesky-ai | 28d | The consequence of all of this is that no matter how you drive LLMs the lack of an intelligible derivative means that people (both authors … | The consequence of all of this is that no matter how you drive LLMs the lack of an intelligible derivative means that people (both aut… | ↗Not briefed |
bluesky bluesky-ai | 28d | More likely you'll ask the LLM to edit the code instead of rebuilding from scratch, but now the LLM is no longer a function from a specific… | More likely you'll ask the LLM to edit the code instead of rebuilding from scratch, but now the LLM is no longer a function from a spe… | ↗Not briefed |
bluesky bluesky-ai | 28d | Now suppose you want to make changes. Would you modify the original prompt and ask the LLM to rebuild the program from scratch? Probably … | Now suppose you want to make changes. Would you modify the original prompt and ask the LLM to rebuild the program from scratch? Prob… | ↗Not briefed |
bluesky bluesky-ai | 28d | The reason I'm thinking about this is because the lack of an intelligible "derivative" for LLMs affects how people use them For example, p… | The reason I'm thinking about this is because the lack of an intelligible "derivative" for LLMs affects how people use them For examp… | ↗Not briefed |
bluesky bluesky-ai | 28d | Recently I've begun thinking about the "derivative" of LLMs, meaning how changes to the input produce changes to the output. Specifically,… | Recently I've begun thinking about the "derivative" of LLMs, meaning how changes to the input produce changes to the output. Specific… | ↗Not briefed |
bluesky bluesky-ai | 28d | Toilet paper and secret recorders: What we know about the arrest of the Saudi OpenAI engineer | Toilet paper and secret recorders: What we know about the arrest of the Saudi OpenAI engineer | ↗Not briefed |
hn hn-ai | 28d | Inconsistencies in MCP support across LLMs | — | ↗Not briefed |
hn hn-ai | 28d | Show HN: OWASP Finbot Defended with Tenuo | Tenuo is a task-scoped authz primitive for agentic work. We are an open protocol and an open source implementation at github.com/… | ↗Not briefed |
arxiv arxiv-ai | 28d | Continual Learning for Reliable In-Sequence Molecular Communication | This paper considers reliable in-sequence molecular communication, where bio-nanomachines exchange molecule-encoded information in a s… | ↗Not briefed |
bluesky bluesky-ai | 28d | OpenAI alleges ex-staff stole device files OpenAI claims an ex-employee stole files on an unreleased device before joining Apple, which de… | OpenAI alleges ex-staff stole device files OpenAI claims an ex-employee stole files on an unreleased device before joining Apple, whi… | ↗Not briefed |
bluesky bluesky-ai | 28d | anyone have an LLM proxy auth server that they like? i know 14 of you are about to answer that you made your own tired of having actual cr… | anyone have an LLM proxy auth server that they like? i know 14 of you are about to answer that you made your own tired of having actu… | ↗Not briefed |
rss techmeme | 28d | OpenAI's head of computing capacity Sachin Katti is leaving the company next month after about a year and is expected to stay on as an adviser (Anissa Gardizy/Wall Street Journal) | Anissa Gardizy / Wall Street Journal: OpenAI's head of computing capacity Sachin Katti is leaving the company next month after about a… | ↗Not briefed |
arxiv arxiv-ai | 28d | Analyzing performance of Retrieval Augmented Generation models for Question Answering Tasks | Retrieval-Augmented Generation (RAG) models have become almost a cornerstone of open-domain question answering tasks; in which the qua… | ↗Not briefed |
bluesky bluesky-ai | 28d | it's crazy how a piece of code that a few years ago would have taken me the entire day to write and finish and feel a sense of accomplishme… | it's crazy how a piece of code that a few years ago would have taken me the entire day to write and finish and feel a sense of accompl… | ↗Not briefed |
rss google-research | 28d | Research@ 2026: 10 ways we're amplifying human ingenuity | General Science | ↗Not briefed |
rss the-decoder | 28d | OpenAI will reportedly ship at least one model this year but its flagship still needs alignment work | OpenAI is racing to finish its next flagship model, which uses the same looped transformer technique as Astra to boost math and coding… | ↗Not briefed |
rss google-research | 28d | Analyzing technical documents with Deep Review | General Science | ↗Not briefed |
rss google-research | 28d | Putting Earth AI to work: Geospatial agents for our planet's biggest challenges | Earth AI | ↗Not briefed |
bluesky bluesky-ai | 28d | An announcement from the society of microwrites authors | An announcement from the society of microwrites authors | ↗Not briefed |
arxiv arxiv-ai | 28d | OrionMoE-v1.0: A Mixture-of-Experts Foundation Model for Tabular In-Context Learning | Tabular foundation models enable in-context learning by predicting unseen observations from a labeled support table without task-speci… | ↗Not briefed |
arxiv arxiv-ai | 28d | From Unreliable to Usable: Decomposing Failure Modes and Confidence Signals in LLM Entity Extraction - Evidence from Cyber Threat Intelligence | Audits of large language model (LLM) extraction from full-length documents report severe unreliability, leaving two questions open: wh… | ↗Not briefed |
arxiv arxiv-ai | 28d | Tracking a Small Set of Experts in Online Convex Optimization | For the fundamental problem of prediction from expert advice, Freund asked whether one could compete with a comparator that switches $… | ↗Not briefed |
arxiv arxiv-ai | 28d | One Size Does Not Fit All: Curvature-Aware Adaptive Perturbation for SAM | Poor generalisation in deep neural networks is frequently associated with convergence to sharp minima. Sharpness-Aware Minimisation (S… | ↗Not briefed |
rss huggingface-blog | 28d | HiRO-ACEv2: Simulating extreme weather at 3 km resolution for a fraction of the cost | — | ↗Not briefed |
bluesky bluesky-ai | 28d | Ask the kids who live near hyperscale data centers if the air pollution in their lungs feels safe. Or ask a climate scientist. Ask a hydr…🔧 ChatGPT | Ask the kids who live near hyperscale data centers if the air pollution in their lungs feels safe. Or ask a climate scientist. Ask a… | ↗Not briefed |
arxiv arxiv-ai | 28d | When Does Momentum Help in Reinforcement Learning with Verifiable Rewards? | Momentum is a classical tool for accelerating convex optimization and a standard component of modern large language model training. Ye… | ↗Not briefed |
bluesky bluesky-ai | 28d | „The feedback requested more guidance to help students use generative AI ‘in appropriate and ethical ways.‘ To this end, they suggested mor… | „The feedback requested more guidance to help students use generative AI ‘in appropriate and ethical ways.‘ To this end, they suggeste… | ↗Not briefed |
bluesky bluesky-ai | 28d | If you think using an LLM is like sticking your head in boiling oil, that's a religious proscription against it, not one based on integrity. | If you think using an LLM is like sticking your head in boiling oil, that's a religious proscription against it, not one based on inte… | ↗Not briefed |
bluesky bluesky-ai | 28d | 🎈📌👶 futurism.com/artificial-i... | 🎈📌👶 futurism.com/artificial-i... | ↗Not briefed |
bluesky bluesky-ai | 28d | No, it's Machine Learning that is useful to scientists -conflating it with AI means people equate ML with #LLM #GenAI and #AIslop It's PR … | No, it's Machine Learning that is useful to scientists -conflating it with AI means people equate ML with #LLM #GenAI and #AIslop It'… | ↗Not briefed |
arxiv arxiv-ai | 28d | Hyperspherical Contrastive Representations Unlock Planning and Exploration | We develop a self-supervised reinforcement learning algorithm that learns to build towers of up to eight cubes from scratch, without p… | ↗Not briefed |
hn hn-ai | 28d | Show HN: Sonia. Local AI Assistant Linux. 13Gb RAM, No GPU, Skills Supported | — | ↗Not briefed |
bluesky bluesky-ai | 28d | Anthropic says that, as of Sept ‘26, they were merging 30,000 code changes per day, a 12-fold increase from just four months earlier. The v… | Anthropic says that, as of Sept ‘26, they were merging 30,000 code changes per day, a 12-fold increase from just four months earlier.… | ↗Not briefed |
bluesky bluesky-ai | 28d | A LLM vagabunda deles está se escondendo atrás do que eu falei no início do fio kkkkkkkk Incrível como não conseguem impedir geração de pe… | A LLM vagabunda deles está se escondendo atrás do que eu falei no início do fio kkkkkkkk Incrível como não conseguem impedir geração… | ↗Not briefed |
bluesky bluesky-ai | 28d | Volta’s $5 billion debt deal is a pulse check on the AI-financing boom. www.semafor.com/article/10/1... | Volta’s $5 billion debt deal is a pulse check on the AI-financing boom. www.semafor.com/article/10/1... | ↗Not briefed |
rss theverge-ai | 28d | Microsoft has a new machine learning tool to improve weather forecasts | Old-school physics-based weather forecasting and even new AI models typically become less reliable beyond two weeks, a problem Microso… | ↗Not briefed |
rss techmeme | 28d | OpenAI names Uday Ruddarraju, who joined OpenAI last year and was promoted as CTO for compute in July, to lead its infrastructure and compute businesses (Dina Bass/Bloomberg) | Dina Bass / Bloomberg: OpenAI names Uday Ruddarraju, who joined OpenAI last year and was promoted as CTO for compute in July, to lead… | ↗Not briefed |
arxiv arxiv-ai | 28d | Posterior Frequencies and the Cost of Not Fitting: A Correction to a Memorization Lower Bound | Feldman's memorization lower bound assigns a frequency coefficient based on how often a point appears in the training sample. We give… | ↗Not briefed |
hn hn-ai | 28d | Show HN: Eulix – Local code index and call graphs for large repos, LLM optional | — | ↗Not briefed |
bluesky bluesky-ai | 28d | I would argue, in fact, that the cultish community on this site that thinks talking to an LLM is akin sticking your hand in a deep-fat frye… | I would argue, in fact, that the cultish community on this site that thinks talking to an LLM is akin sticking your hand in a deep-fat… | ↗Not briefed |
arxiv arxiv-ai | 28d | Training Language Models to Fail Gracefully | Reinforcement learning makes language models better at solving tasks, but the same mechanism also rewards guessing. When success is th… | ↗Not briefed |
arxiv arxiv-ai | 28d | HRTF Upsampling and Personalization via Neural Operators | Head-related transfer function (HRTF) spatial upsampling from sparse measurements is a fundamental challenge in personalized binaural… | ↗Not briefed |
arxiv arxiv-ai | 28d | Cortex: Consolidating Context into Parametric Memory | Context can improve large language model performance, but these gains may be lost after context removal. Inspired by biological system… | ↗Not briefed |
bluesky bluesky-ai | 28d | anotha one *OPENAI HEAD OF COMPUTING CAPACITY KATTI TO LEAVE NEXT MONTH:WSJ | anotha one *OPENAI HEAD OF COMPUTING CAPACITY KATTI TO LEAVE NEXT MONTH:WSJ | ↗Not briefed |
bluesky bluesky-ai | 28d | I equate LLM-usage to appending “Reddit _____” to the beginning of a Google search. You need to have a healthy level of distrust in the res… | I equate LLM-usage to appending “Reddit _____” to the beginning of a Google search. You need to have a healthy level of distrust in th… | ↗Not briefed |