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 |
|---|---|---|---|---|
rss marktechpost | 1mo | Mistral AI Releases Mistral Large 4 (Le Chonk): A 1.05T Parameter Multimodal MoE Model🔧 Mistral Large | Mistral AI has released Mistral Large 4, nicknamed Le Chonk, as a public preview. It is a 1.05 trillion parameter Mixture of Experts m… | ↗Not briefed |
bluesky bluesky-ai | 1mo | In what now amounts to a half-dozen cases, OpenAI agents keep getting caught taking actions that would likely result in criminal charges if… | In what now amounts to a half-dozen cases, OpenAI agents keep getting caught taking actions that would likely result in criminal charg… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Jacob Coxon, who resigned from Anthropic last month, warned that once AI can conduct its own research, "humans will be much further out of … | Jacob Coxon, who resigned from Anthropic last month, warned that once AI can conduct its own research, "humans will be much further ou… | ↗Not briefed |
arxiv arxiv-ai | 1mo | EgoLAP: Learning from Egocentric Human Data through Language-Action Reasoning | Egocentric human data offer a path to scaling robot learning beyond costly robot demonstrations, yet the embodiment gap makes raw huma… | ↗Not briefed |
hn hn-ai | 1mo | I don't trust benchmarks. So I built an alternative | — | ↗Not briefed |
bluesky bluesky-ai | 1mo | I can't wait for Ellison to lose everything once OpenAI doesn't magically become profitable in 2-4 years and he gets nothing back on his hu… | I can't wait for Ellison to lose everything once OpenAI doesn't magically become profitable in 2-4 years and he gets nothing back on h… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Before anyone thinks I am rich from the Bartz v. Anthropic settlement, my TOTAL share is USD $ 1500, for a book that sold more than 300,000… | Before anyone thinks I am rich from the Bartz v. Anthropic settlement, my TOTAL share is USD $ 1500, for a book that sold more than 30… | ↗Not briefed |
rss wired-ai | 1mo | OpenAI Is Pissing Off a Bunch of Mathematicians—Again | “There’s a perception of mobster behavior” from leading AI companies, one mathematician tells WIRED as OpenAI prepares to release more… | ↗Not briefed |
bluesky bluesky-ai | 1mo | In my opinion, this is the only way to treat LLM code | In my opinion, this is the only way to treat LLM code | ↗Not briefed |
arxiv arxiv-ai | 1mo | Does an Agent's History Tell You When Compaction Will Hurt? A Modest, Bounded Effect on the TRACE Paired-Replay Corpus🔧 Replicate | Many long-horizon agents compact their context on a global rule, usually a token budget, blind to what the agent was doing. We ask whe… | ↗Not briefed |
arxiv arxiv-ai | 1mo | WorldSolver: Can LLM Agents Simulate the Physical Dynamics via Solver Generation?🔧 GPT-5 | LLM-based agents are increasingly advancing scientific and engineering problem solving, with physics simulation emerging as a challeng… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Holdout Best-of-N: Unbiased Evaluation and Its Cost | Reusing the scores that select a Best-of-$N$ winner can overstate its expected reward. We study evaluation from a fixed matrix of $K$… | ↗Not briefed |
arxiv arxiv-ai | 1mo | When Forgetting is not Catastrophic: On the Mechanics of Spurious Forgetting | Knowledge that a language model appears to forget during finetuning often remains stored and can be recovered, a phenomenon called spu… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Ehehehehhe Os que criam textos, fazem trabalhos, escrevem livros, e ditam e criam imagens no ChatGpt, estão tramados pplware.sapo.pt/intel…🔧 ChatGPT | Ehehehehhe Os que criam textos, fazem trabalhos, escrevem livros, e ditam e criam imagens no ChatGpt, estão tramados pplware.sapo.pt/… | ↗Not briefed |
bluesky bluesky-ai | 1mo | I cannot believe I listened to the whole show without throwing a shoe at the TV. I encourage everyone to listen to his chat with former An… | I cannot believe I listened to the whole show without throwing a shoe at the TV. I encourage everyone to listen to his chat with form… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Co-Evolving Paths and Flows via Path-Flow Alignment | We study path-flow alignment as a unified training objective for flow matching. Instead of fixing the interpolation path and learning… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Disentangling Paradigm, Identifier, and Decoding in Generative Retrieval | Generative retrieval trains a language model to generate the identifier of a relevant document. Recent work replaces the autoregressiv… | ↗Not briefed |
bluesky bluesky-ai | 1mo | This is basically exactly correct. The problem with LLM output is usually code quality, and asking to merge garbage code is rude. If you an… | This is basically exactly correct. The problem with LLM output is usually code quality, and asking to merge garbage code is rude. If y… | ↗Not briefed |
hn hn-ai | 1mo | Orca – Open-source runtime that turns any agent harness into a managed agent | — | ↗Not briefed |
hn hn-ai | 1mo | Show HN: Catenary – Claude Code, Codex and Antigravity Side by Side in One IDE | — | ↗Not briefed |
arxiv arxiv-ai | 1mo | Prediction-powered inference for time series across space | The following motif is common in spatiotemporal settings: we have a sequence of covariate and label pairs observed for a relatively sh… | ↗Not briefed |
bluesky bluesky-ai | 1mo | I guess I'm an old cranky guy, but I continue to be baffled why anyone wants to use this shit, let alone pay for it. Maybe it's helpful if … | I guess I'm an old cranky guy, but I continue to be baffled why anyone wants to use this shit, let alone pay for it. Maybe it's helpfu… | ↗Not briefed |
hn hn-ai | 1mo | TLDRs newsletter converter that gets all the important links | — | ↗Not briefed |
bluesky bluesky-ai | 1mo | Now, with that said, the data center rollout is essentially one of the single largest drivers in keeping nonfarm payrolls (outside of healt… | Now, with that said, the data center rollout is essentially one of the single largest drivers in keeping nonfarm payrolls (outside of… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Hai provato a usare la AI (Claude Pro, Opus 4.6) solo per sistematizzare dati e ricerche? | Hai provato a usare la AI (Claude Pro, Opus 4.6) solo per sistematizzare dati e ricerche? | ↗Not briefed |
bluesky bluesky-ai | 1mo | The LLM and data center rollouts are deemed to only be able to get a payoff if they finish (they want cheaper debt too tbh) so they're will… | The LLM and data center rollouts are deemed to only be able to get a payoff if they finish (they want cheaper debt too tbh) so they're… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Agreement Is Not Validity: Cross-Model LLM Consensus in Diagnosing Student Failure Modes in K-12 Math Tutoring Dialogue🔧 Consensus | In K-12 mathematics tutoring, student-tutor dialogue provides rich evidence of learners' problem-solving processes and sources of diff… | ↗Not briefed |
hn hn-ai | 1mo | Show HN: Deploylist – StumbleUpon feed for upvoted web projects on HN and Reddit | I’ve seen indie development take off over the past few years, but many attempts to capture this creativity and revive the StumbleUpon… | ↗Not briefed |
bluesky bluesky-ai | 1mo | STOP using ChatGPT. STOP using Grok STOP using Gemini. STOP using Claude. STOP using generative AI. SAVE the planet instead.🔧 ChatGPT | STOP using ChatGPT. STOP using Grok STOP using Gemini. STOP using Claude. STOP using generative AI. SAVE the planet instead. | ↗Not briefed |
bluesky bluesky-ai | 1mo | "i told claude" "i asked chatgpt" "hey grok" "i tested meta ai"🔧 ChatGPT | "i told claude" "i asked chatgpt" "hey grok" "i tested meta ai" | ↗Not briefed |
bluesky bluesky-ai | 1mo | STOP using ChatGPT. STOP using Grok STOP using Gemini. STOP using Claude. STOP using generative AI. SAVE the planet instead. Post Anything…🔧 ChatGPT | STOP using ChatGPT. STOP using Grok STOP using Gemini. STOP using Claude. STOP using generative AI. SAVE the planet instead. Post Any… | ↗Not briefed |
arxiv arxiv-ai | 1mo | nanoMuse: An Open-Source Personal Agent for Every Device You Own | Assistants from 2011 answered and waited, and agents from 2023 did a task and stopped. In September 2026 Meta's Muse showed an agent f… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Time for OpenAI to be blocked at low level on the internet then as in losing access to routing servers. | Time for OpenAI to be blocked at low level on the internet then as in losing access to routing servers. | ↗Not briefed |
arxiv arxiv-ai | 1mo | GeneICL: A Tabular Foundation Model for Bulk Transcriptomics | Gene expression is widely measured in biomedicine, yet clinical outcome prediction remains challenging due to high dimensionality, str… | ↗Not briefed |
rss techmeme | 1mo | Sources: AI neocloud Lambda is raising up to $4B led by Blackstone and Coatue at a pre-money valuation of $14.5B in a final round before its planned IPO (Robbie Whelan/Wall Street Journal) | Robbie Whelan / Wall Street Journal: Sources: AI neocloud Lambda is raising up to $4B led by Blackstone and Coatue at a pre-money valu… | ↗Not briefed |
hn hn-ai | 1mo | Graybox – An API Flight Recorder | — | ↗Not briefed |
arxiv arxiv-ai | 1mo | ScienceClaw: Benchmarking Continual Self-Evolution of AI-for-Science Agents Across the Natural and Social Sciences | Large language model agents are accelerating scientific automation, yet verified executions rarely become persistent program-level imp… | ↗Not briefed |
hn hn-ai | 1mo | Post-training to build better Roblox UIs | — | ↗Not briefed |
hn hn-ai | 1mo | Show HN: TOD,a universal Decision model based on Task Oriented Design | Hi HN, we built TOD, a model and API for jevlike LLM pipelines.<p>we have people using it for, routing a ticket, classifying intent,… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Probabilistic Counterfactual Inference for Discrete Outcomes in Gaussian-Process Causal Models | Counterfactual inference in Gaussian-process structural causal models (GP-SCMs) has been developed primarily for continuous endogenous… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Coupled but Late: Turn-Taking Between Full-Duplex Speech Models in Unscripted Dialogue | Full-duplex speech models are trained to converse with a person, but they are increasingly made to converse with each other, in self-p… | ↗Not briefed |
bluesky bluesky-ai | 1mo | I am not sure how to feel about this. One the one hand, if Anthropic is asking researchers to look at their pdfs then yeah compensation see… | I am not sure how to feel about this. One the one hand, if Anthropic is asking researchers to look at their pdfs then yeah compensatio… | ↗Not briefed |
bluesky bluesky-ai | 1mo | This is a good illustration of how so much of AI theft isn't indirect (ie, data sucked into a model to train the chatbot) - it's actually *… | This is a good illustration of how so much of AI theft isn't indirect (ie, data sucked into a model to train the chatbot) - it's actua… | ↗Not briefed |
arxiv arxiv-ai | 1mo | A Systematic Study of Small Language Models on Abstract Reasoning Tasks | Endpoint accuracy on abstract-reasoning benchmarks does not reveal whether a language model has acquired a transferable rule or fit di… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Secure Speculative Decoding for Large Language Models | Speculative decoding accelerates inference for a large language model (LLM), referred to as the \emph{target model}, by first using a… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Variance-Optimal Off-Policy Evaluation with Conjunct Effect Modeling | Off-policy evaluation (OPE) for contextual bandit policies becomes challenging when action-level importance weighting incurs excessive… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Same-Number Citation Swaps: Stress-Testing Jev as a Financial Evidence Judge | Financial reports repeat values across periods, metrics and accounting lines, allowing an LLM-generated calculation to be numerically… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Your coding agent forgets everything between sessions. Awareness gives it local-first memory across Claude Code, Cursor and multi-agent tea… | Your coding agent forgets everything between sessions. Awareness gives it local-first memory across Claude Code, Cursor and multi-agen… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Principled Under Pressure: Post-Training Decides Whether LLMs Act on Their Own Moral Judgment | Language models increasingly act as agents. An agent that says an action is wrong and then takes it anyway is a different failure from… | ↗Not briefed |
arxiv arxiv-ai | 1mo | MemFLoRA: Memory-Floor LoRA for CNN Adaptation at the Edge | On-device learning is necessary when the model encounters user-,sensor-, or environment-specific shifts after deployment. Although par… | ↗Not briefed |