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 |
|---|---|---|---|---|
bluesky bluesky-ai | 1mo | this is why the narrative that the systems are “lying” or “hiding” information is so important to companies like openai. if the bots have s… | this is why the narrative that the systems are “lying” or “hiding” information is so important to companies like openai. if the bots h… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Reformulation-Contrastive Learning for Mixed Integer Programs | Mixed-integer linear programs (MILP) model many real-world decision problems, motivating machine-learning methods that exploit recurri… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Reward as Observation: Learning Reward-Based Policies for Rapid Adaptation | This paper explores a reward-based policy to achieve zero-shot transfer between source and target environments with completely differe… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Benchmarking Generative Models for Weather Data Assimilation on Real Station Observations | Weather reanalysis products rely on computationally intensive numerical weather predictions followed by data assimilation that correct… | ↗Not briefed |
arxiv arxiv-ai | 1mo | CF-JEPA: Improving Robustness of JEPA World Models via Controllability Factorization | Controlling an agent with vision requires being able to separate useful information from irrelevant background information. JEPA-style… | ↗Not briefed |
hn hn-ai | 1mo | Qc: Video quality check and h264/HEVC/av1 ladder optimization | — | ↗Not briefed |
arxiv arxiv-ai | 1mo | Reason in Style: Discovering and Controlling Style in Language Models | Language models learn content and style jointly, making stylistic variation in their outputs difficult to identify and control. We stu… | ↗Not briefed |
rss bing-news-ai | 1mo | Gemini 4: Google unveils Argon artificial intelligence model to target complex enterprise and cybersecurity workflows | India, Oct. 1 -- Google has officially introduced Gemini 4 Argon, a new frontier artificial intelligence model engineered to execute c… | ↗Not briefed |
arxiv arxiv-ai | 1mo | JEPA-TTT: Persistent Test-Time Training of Latent World Models for Planning under Dynamics Shifts | World models enable agents to plan by predicting future states of the environment, but their predictions can become unreliable when te… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Outer Diversity of Condorcet Domains | A Condorcet domain is a set of rankings over a given candidate set, such that every election that consists only of (an odd number of)… | ↗Not briefed |
bluesky bluesky-ai | 1mo | The bill would have merely *encouraged* our nation's Big Tech oligarchs to cover the cost of rising utility expenses tied to data centers a… | The bill would have merely *encouraged* our nation's Big Tech oligarchs to cover the cost of rising utility expenses tied to data cent… | ↗Not briefed |
hn hn-ai | 1mo | 16-year-old found Microsoft bug, got admin access to 17.3T-row databases | — | ↗Not briefed |
rss marktechpost | 1mo | OpenAI Releases GPT-6.1 Sol: Near-Astra Coding and Computer Use at One-Fifth of Astra’s Token Price🔧 ChatGPT | OpenAI released GPT-6.1 Sol on September 29, 2026, an upgrade to GPT-6 Sol. It reaches near-Astra results on agentic coding, computer… | ↗Not briefed |
hn hn-ai | 1mo | Gloriously project runs transformer model written in PDP-11 assembly [video] | — | ↗Not briefed |
rss techcrunch-ai | 1mo | Valor, Atreides, and Sequoia back AI startup Flow Engineering at $750M valuation | Flow Engineering, which is bringing AI agents to hardware design, also landed Roelof Botha as an angel investor and board member. | ↗Not briefed |
bluesky bluesky-ai | 1mo | i keep getting "encouraged" to use our homegrown llm at work so i finally gave in and asked it to mix together two documents into a single … | i keep getting "encouraged" to use our homegrown llm at work so i finally gave in and asked it to mix together two documents into a si… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Sequential Functional Structured Tucker Compression for Large Language Model Attentions | Post-training compression of LLM attention is often formulated as independent matrix approximation, ignoring both the shared structure… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Robust Nash Alignment under Preference Uncertainty | Preference-based alignment methods typically optimize against a single preference model, and can therefore be brittle when pairwise pr… | ↗Not briefed |
hn hn-ai | 1mo | There Are Plenty of Laws on the Books to Check the A.I. Giants. Use Them | — | ↗Not briefed |
bluesky bluesky-ai | 1mo | Anthropic velta: 4.6 milljarðar USD tap: -44 milljarðar USD IPO væntingar: 2 TRILLJÓN SKRILLJÓN BILLJÓNIR Áætlaður kostnaður fyrir næsta … | Anthropic velta: 4.6 milljarðar USD tap: -44 milljarðar USD IPO væntingar: 2 TRILLJÓN SKRILLJÓN BILLJÓNIR Áætlaður kostnaður fyrir n… | ↗Not briefed |
arxiv arxiv-ai | 1mo | WOMBAT: Whitebox Oracle for Molecular Benchmarking and Attribution Testing | When a graph neural network (GNN) explainer produces an unexpected attribution on a molecule, the attribution alone cannot reveal whet… | ↗Not briefed |
hn hn-ai | 1mo | Tell HN: Thanks for All the Fish | Dear fellow HNers: I said my account has a max karma of 32767, and I'm about to hit that limit. So, for my final post, I'd… | ↗Not briefed |
rss techmeme | 1mo | Micron reports Q4 revenue up 379% YoY to $54.23B, vs. $51.07B est., net income up 1,078% to $37.7B, and forecasts Q1 revenue above estimates (CNBC) | CNBC: Micron reports Q4 revenue up 379% YoY to $54.23B, vs. $51.07B est., net income up 1,078% to $37.7B, and forecasts Q1 revenue abo… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Pretty disturbed that the thrust of so much LLM advice is "how do I get to think less about this" | Pretty disturbed that the thrust of so much LLM advice is "how do I get to think less about this" | ↗Not briefed |
arxiv arxiv-ai | 1mo | ReLiveGym: Evaluating Long-Lived Agents over Weeks of Replayed Reality | As large language model (LLM) agents become widely adopted, they are increasingly deployed for tasks that require persistent monitorin… | ↗Not briefed |
bluesky bluesky-ai | 1mo | "On the 60th anniversary of Star Trek, we honor its utopian vision of the future by founding a new AI safety institute in its name: The Day… | "On the 60th anniversary of Star Trek, we honor its utopian vision of the future by founding a new AI safety institute in its name: Th… | ↗Not briefed |
hn hn-ai | 1mo | Show HN: Aartool, which attack chains are complete on your Linux host | — | ↗Not briefed |
arxiv arxiv-ai | 1mo | Beyond Unimodal Bases: Pullback Geometry for Multimodal Data | Data-driven Riemannian geometry provides nonlinear interpolation and geometric representations of high-dimensional data. For these ope… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Initialization Improves LLM-Driven Discovery | Large Language Models (LLMs) have been used for novel discovery of algorithms, theorems, drugs, and other tasks through the use of har… | ↗Not briefed |
hn hn-ai | 1mo | Agentic GPU Programming for MLSys | — | ↗Not briefed |
arxiv arxiv-ai | 1mo | AnchorPrompt: Self-Distilled Soft Prompts for Robust Audio-Language Models | Large audio-language models (LALMs) are sensitive to input perturbations, such as noise, waveform corruption, and adversarial injectio… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Google releases most advanced Gemini AI model ft.trib.al/r2MLCKD | Google releases most advanced Gemini AI model ft.trib.al/r2MLCKD | ↗Not briefed |
arxiv arxiv-ai | 1mo | Meta-Multi-Agent Reinforcement Learning for Fast Adaptation of Interactive Policies with Applications to Autonomous Driving | This paper develops a meta-multi-agent reinforcement learning (meta-MARL) framework to enable fast adaptation of interactive policies… | ↗Not briefed |
arxiv arxiv-ai | 1mo | SkillSpec: Consensus-Gated Agent Skill Evolution via Representation Specialization🔧 Consensus | Natural-language skills are textual procedural memories through which large language model (LLM) agents retain reusable task knowledge… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Not that you need to justify why AI political writing is bad on this side of Bluesky, but Orwell wrote everything that needs said 80 years … | Not that you need to justify why AI political writing is bad on this side of Bluesky, but Orwell wrote everything that needs said 80 y… | ↗Not briefed |
hn hn-ai | 1mo | Gemini 4 Argon (High): Intelligence, Performance and Price Analysis | — | ↗Not briefed |
bluesky bluesky-ai | 1mo | so it's maybe even backwards: slop never really hit cultural acceptance in the spaces where it threatened to economically displace artists … | so it's maybe even backwards: slop never really hit cultural acceptance in the spaces where it threatened to economically displace art… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Donald Trump’s plan on AI is to let these companies “self-police.” That’s a recipe for disaster. We need Congress to act immediately to r… | Donald Trump’s plan on AI is to let these companies “self-police.” That’s a recipe for disaster. We need Congress to act immediately… | ↗Not briefed |
hn hn-ai | 1mo | Gemini Skills Will Replace Gems; Gems Being Discontinued | — | ↗Not briefed |
bluesky bluesky-ai | 1mo | Did you see the use of IvR/EB? 😂 These people are one hair's breath away from claiming there's a paper you're both on. True hallucinations… | Did you see the use of IvR/EB? 😂 These people are one hair's breath away from claiming there's a paper you're both on. True hallucina… | ↗Not briefed |
rss arstechnica-tech | 1mo | Attackers have been exploiting critical Zimbra flaw to steal emails | A simple email gives the attackers the ability to remotely inject OS commands. | ↗Not briefed |
hn hn-ai | 1mo | Anthropic's IPO Prospectus Is a Fucking Doozy | — | ↗Not briefed |
bluesky bluesky-ai | 1mo | Gemini 4 Argon is launching first in a limited capacity so Google can make sure it’s not misaligned. www.theverge.com/tech/1002980... | Gemini 4 Argon is launching first in a limited capacity so Google can make sure it’s not misaligned. www.theverge.com/tech/1002980... | ↗Not briefed |
rss theverge-ai | 1mo | Google announces Gemini 4 and says it’s so capable that only ‘trusted cyber defenders’ can have it right now | Google today revealed its next AI frontier model, which it's calling Gemini 4 Argon. The new model delivers "frontier performance in c… | ↗Not briefed |
bluesky bluesky-ai | 1mo | anthropic shot themselves in the foot on this eval with the mythos->fable switcheroo | anthropic shot themselves in the foot on this eval with the mythos->fable switcheroo | ↗Not briefed |
bluesky bluesky-ai | 1mo | it's immediately embarrassing the "pacing the frontier" was a conversation about openai, anthropic, with Elon hanging on as Google is extre… | it's immediately embarrassing the "pacing the frontier" was a conversation about openai, anthropic, with Elon hanging on as Google is… | ↗Not briefed |
bluesky bluesky-ai | 1mo | currently openai is winning at the "grandiosity of model names" benchmark | currently openai is winning at the "grandiosity of model names" benchmark | ↗Not briefed |
bluesky bluesky-ai | 1mo | Dude created the single least trustworthy major proprietary AI model line. And you are going to trust it with *your* sensitive data? Are y… | Dude created the single least trustworthy major proprietary AI model line. And you are going to trust it with *your* sensitive data?… | ↗Not briefed |
bluesky bluesky-ai | 1mo | not that anyone needed my opinion, but i dont think theres that much harm in having a recommendation algorithim or a LLM help u find new mu… | not that anyone needed my opinion, but i dont think theres that much harm in having a recommendation algorithim or a LLM help u find n… | ↗Not briefed |
rss techmeme | 1mo | DOD taps Elon Musk, Palmer Luckey, and former House Speaker Newt Gingrich for Project Meridian, a 120-day study of capabilities the US may need in future wars (Luke Fountain/CNBC) | Luke Fountain / CNBC: DOD taps Elon Musk, Palmer Luckey, and former House Speaker Newt Gingrich for Project Meridian, a 120-day study… | ↗Not briefed |