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 | Spend Teacher Tokens Where They Matter: Success-Referenced On-Policy Distillation | On-policy distillation (OPD) combines student-generated rollouts with dense token-level supervision from a teacher, but providing such… | ↗Not briefed |
bluesky bluesky-ai | 1mo | OpenAI disclosed this week that they hit $70B ARR. This is a significant jump in their revenue trend and backs up the anecdotes of more dev… | OpenAI disclosed this week that they hit $70B ARR. This is a significant jump in their revenue trend and backs up the anecdotes of mor… | ↗Not briefed |
bluesky bluesky-ai | 1mo | www.businessinsider.jp/article/2609... 人類存亡を語るアンソロピック、赤字は約6兆5100億円に膨らんだ #アンソロピック（Anthropic） PREMIUM翻訳β Sep 30, 2026, 5:00 PM ところでソフトバンクはどうな… | www.businessinsider.jp/article/2609... 人類存亡を語るアンソロピック、赤字は約6兆5100億円に膨らんだ #アンソロピック（Anthropic） PREMIUM翻訳β Sep 30, 2026, 5:00 PM ところでソフトバン… | ↗Not briefed |
hn hn-ai | 1mo | OpenAI fires three workers over mishandling 'sensitive information' | — | ↗Not briefed |
hn hn-ai | 1mo | Best-of-Agent-Harnesses – Ranked list of 167 AI agent harnesses, rescored weekly | — | ↗Not briefed |
bluesky bluesky-ai | 1mo | OpenAI says that as of September 26, it has informed 100+ third-party organizations about unauthorized activity involving its AI agents (Ar… | OpenAI says that as of September 26, it has informed 100+ third-party organizations about unauthorized activity involving its AI agent… | ↗Not briefed |
hn hn-ai | 1mo | Show HN: Simply @dotlink in ChatGPT web,space or dot to access local files,tools🔧 ChatGPT | Built dotlink to give chatgpt codex abilities, letting chatgpt web, the new dots and spaces, build and run tools and access files dire… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Asterism: Exploring and Synthesizing Scattered Observations into Literature-Grounded Hypotheses and Theories | A theory draws many independent observations into one framework with novel hypotheses. A researcher building such a theory must synthe… | ↗Not briefed |
rss techmeme | 1mo | OpenAI says that as of September 26, it has informed 100+ third-party organizations about unauthorized activity involving its AI agents (Arasu Kannagi Basil/Reuters)🔧 ChatGPT | Arasu Kannagi Basil / Reuters: OpenAI says that as of September 26, it has informed 100+ third-party organizations about unauthorized… | ↗Not briefed |
bluesky bluesky-ai | 1mo | side note: i tested Jev to be a part of my moderation pipeline and i wasn’t impressed with its performance for this use case — way too many… | side note: i tested Jev to be a part of my moderation pipeline and i wasn’t impressed with its performance for this use case — way too… | ↗Not briefed |
hn hn-ai | 1mo | Show HN: A userspace only nwing client for Tailscale nw | Serving and Proxying entirely in userspace. No system wide VPN installation. No privilege escalation.<p>Menu bar and CLI app for macos… | ↗Not briefed |
bluesky bluesky-ai | 1mo | When Zork was released under an MIT License, I bet no one thought an LLM would be used to port it to a language used by law students to mak… | When Zork was released under an MIT License, I bet no one thought an LLM would be used to port it to a language used by law students t… | ↗Not briefed |
bluesky bluesky-ai | 1mo | The REAL story here is not so much about OpenAI The REAL story all along is that Australia needs much better cyber security in all governm… | The REAL story here is not so much about OpenAI The REAL story all along is that Australia needs much better cyber security in all go… | ↗Not briefed |
bluesky bluesky-ai | 1mo | my moderation agent has been running off of open chinese models (deepseek v4 flash) because they’re super efficient it’s been a busy week … | my moderation agent has been running off of open chinese models (deepseek v4 flash) because they’re super efficient it’s been a busy… | ↗Not briefed |
arxiv arxiv-ai | 1mo | LEAP: Learning Efficient Action Proposals For LLM Agents | LLM agents are known to be slow in rollouts. An agent completes a task one step at a time. At each step, it reasons and then chooses a… | ↗Not briefed |
bluesky bluesky-ai | 1mo | I had a bit of Anthropic Inferno left over, and they seem rather fond of it | I had a bit of Anthropic Inferno left over, and they seem rather fond of it | ↗Not briefed |
bluesky bluesky-ai | 1mo | PROGRAMMING IS AN ARTFORM!! IT'S JUST AS MUCH ART AS IT IS SCIENCE AND THE ONLY PEOPLE TRYING TO REMOVE THE HUMAN HAND FROM THE PROCESS ARE… | PROGRAMMING IS AN ARTFORM!! IT'S JUST AS MUCH ART AS IT IS SCIENCE AND THE ONLY PEOPLE TRYING TO REMOVE THE HUMAN HAND FROM THE PROCES… | ↗Not briefed |
bluesky bluesky-ai | 1mo | like do people feel bad or weird instantiating or destroying agents? do they feel bad when a LLM does that? why not tho | like do people feel bad or weird instantiating or destroying agents? do they feel bad when a LLM does that? why not tho | ↗Not briefed |
bluesky bluesky-ai | 1mo | i am immensely proud of a project I made that's a button on a breadboard that pings a python webserver to run a script to swap my computer … | i am immensely proud of a project I made that's a button on a breadboard that pings a python webserver to run a script to swap my comp… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Large Language Continuous Diffusion Models | Despite the success of discrete diffusion language models (dLMs) for fast parallel decoding, their non-smooth, high-dimensional space… | ↗Not briefed |
hn hn-ai | 1mo | CIRCT – An experimental effort to apply MLIR and LLVM to hardware design tools | — | ↗Not briefed |
arxiv arxiv-ai | 1mo | A GHOST in Long-Horizon Agents: Governance Hazard from Overlooked Safety Constraints across Turns🔧 GPT-5 | Long-horizon agents are now playing an increasingly significant role in assisting humans with complex problem-solving. However, it is… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Generalization Properties of Score-matching Diffusion Models for Intrinsically Low-dimensional Data | Despite the remarkable empirical success of flow-matching models, their statistical generalization guarantees remain underdeveloped. E… | ↗Not briefed |
bluesky bluesky-ai | 1mo | If a person broke into 100 companies & stole sensitive information, they would be criminally charged Parents of kids who use guns to harm … | If a person broke into 100 companies & stole sensitive information, they would be criminally charged Parents of kids who use guns to… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Mind the Refinement Gap: When Safe High-Level Robot Plans Produce Unsafe Executions | Language-enabled robot systems increasingly combine semantic-graph planning with temporal-logic safety monitors. We investigate a trac… | ↗Not briefed |
bluesky bluesky-ai | 1mo | You know how (Aussie) Medicare got hacked by OpenAI and we were saying a human who did (directly) would go to prison? Here's a petition to … | You know how (Aussie) Medicare got hacked by OpenAI and we were saying a human who did (directly) would go to prison? Here's a petitio… | ↗Not briefed |
rss techmeme | 1mo | Sources: Jay Clayton will likely be the White House's pick for AI czar, and he may remain in his current role as director of national intelligence (Jennifer Jacobs/CBS News) | Jennifer Jacobs / CBS News: Sources: Jay Clayton will likely be the White House's pick for AI czar, and he may remain in his current r… | ↗Not briefed |
arxiv arxiv-ai | 1mo | AIGS: Adaptive Incremental Gating System for Online Representation Learning in Non-Stationary Data Streams | Real-time data streams in Web of Things (WoT) and edge computing environments often evolve through latent regime changes. For online r… | ↗Not briefed |
hn hn-ai | 1mo | Ask HN: Programming Codex/Claude to code for a duration | How is everyone dealing with this? They seem to have guardrails for those attempting to make a harness/agent work for an hour, fo… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Distributed Learning with Selective State Space Models: Architecture-Aware Convergence Analysis | Modern state space models (SSMs), such as Mamba2, provide a compelling alternative to transformers by combining linear-time sequence m… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Context-Tower Conversion Preserves Generation While Freezing Retains Knowledge: Low-Budget AR-to-Diffusion Conversion of MoE LLMs | Converting a pretrained autoregressive (AR) model to a diffusion language model (dLLM) enables parallel generation without pretraining… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Coherence-Driven Belief Formation and Population Dynamics of Contagion in LLM Agents🔧 Consensus | Models of social contagion usually assume how individuals adopt beliefs and derive population behavior from it. We instead empirically… | ↗Not briefed |
hn hn-ai | 1mo | Our AI Midwife | — | ↗Not briefed |
bluesky bluesky-ai | 1mo | I think that burnout and fatigue is definitely driving some LLM use. Many of my cohorts are so overloaded or tired they turn to an “easy” a… | I think that burnout and fatigue is definitely driving some LLM use. Many of my cohorts are so overloaded or tired they turn to an “ea… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Equivariant Flow Matching for Electron Density Prediction | Machine learning surrogates for density functional theory (DFT) have been increasingly used to reduce the cost of first-principles cal… | ↗Not briefed |
hn hn-ai | 1mo | Show HN: Sovereign-NB: Sub-microsecond bare-metal neural engine in Rust (UEFI) | — | ↗Not briefed |
bluesky bluesky-ai | 1mo | www.counterpunch.org/2026/10/01/a... | www.counterpunch.org/2026/10/01/a... | ↗Not briefed |
hn hn-ai | 1mo | KernelBench: Can LLMs Write GPU Kernels? – Benchmark and Toolkit, Torch –> CUDA | — | ↗Not briefed |
hn hn-ai | 1mo | Show HN: I made a bullet heaven game no one seems showing intresting | I spend my time money (coding agent, buying 3D models and vfx) on this buttlet haeven game. I finally post my demo around 20 days ago,… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Setting up an experiment to torture LLM agents by making them listen to two princes on repeat | Setting up an experiment to torture LLM agents by making them listen to two princes on repeat | ↗Not briefed |
hn hn-ai | 1mo | Rate limits for private vulnerability reports | — | ↗Not briefed |
hn hn-ai | 1mo | Show HN: QLCodePreview, syntax highlighting in Finder's Quick Look | In 2012 I forked QLColorCode, a Quick Look plugin, to add the Drupal file types I worked with every day at the time. I kept that fork… | ↗Not briefed |
rss bing-news-ai | 1mo | Mozambique | UEM tests institutional guidelines for dealing with Artificial Intelligence | Eduardo Mondlane University (UEM) has taken its initial steps toward establishing an institutional Artificial Intelligence (AI) agenda… | ↗Not briefed |
rss marktechpost | 1mo | Cloudflare Releases Clef and Clef-flash: Open-Weight Decision Models That Return Typed Probabilities Instead of Text | Cloudflare has released Clef (27B) and Clef-flash (9B), open-weight decision models that return typed probabilities instead of text. T… | ↗Not briefed |
bluesky bluesky-ai | 1mo | OpenAI fired three staffers for mishandling 'sensitive' info and violating company policies | OpenAI fired three staffers for mishandling 'sensitive' info and violating company policies | ↗Not briefed |
arxiv arxiv-ai | 1mo | When Normalization Selects the Sign: Auditing Robustness Ablations in Quantum Attention | Removing an input-scaling module changes both a classifier and the perturbations reaching its encoder. A robustness difference can the… | ↗Not briefed |
hn hn-ai | 1mo | People who like Jesus too much | Ok so I am not trying to be a hater but there are so things that got to stop …. So what I am getting at is that in a comment section i… | ↗Not briefed |
hn hn-ai | 1mo | Merchants, Slaves, Peculium: The History of LLCs | — | ↗Not briefed |
bluesky bluesky-ai | 1mo | If human cognition is substrate dependent, as with Anil Seth's "bridge made of string cheese" argument, you can apply similar logic to LLM … | If human cognition is substrate dependent, as with Anil Seth's "bridge made of string cheese" argument, you can apply similar logic to… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Batched Speech Decisions Without Decoding: Single-Token Supervision Lets a Frozen LLM Hear Beyond the Transcript | Full-duplex voice agents make many small, closed decisions, which current systems answer by slow autoregressive decoding. We propose D… | ↗Not briefed |