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 |
|---|---|---|---|---|
hn hn-ai | 2mo | Show HN: Trendradar for Horizon Scanning and Foresight | This is an update to our old Trendradar (<a href="https://news.ycombinator.com/item?id=45332076">https://news… | ↗Not briefed |
arxiv arxiv-ai | 2mo | Personalized Federated Learning through Global Knowledge Distillation and Local Head Adaptation | Statistical heterogeneity limits federated learning when a single global classifier cannot represent client-specific label distributio… | ↗Not briefed |
bluesky bluesky-ai | 2mo | Also "It creates wild technical debt" like it's somehow impossible to read code when an llm generates it | Also "It creates wild technical debt" like it's somehow impossible to read code when an llm generates it | ↗Not briefed |
bluesky bluesky-ai | 2mo | Mike Johnson on OpenAI and Anthropic: "They can self-police. They can self-regulate." | Mike Johnson on OpenAI and Anthropic: "They can self-police. They can self-regulate." | ↗Not briefed |
bluesky bluesky-ai | 2mo | I lurk (with popcorn) on the Wikipedia noticeboard dealing with LLM use: with great regularity, LLM edits have sourcing errors: not just ha… | I lurk (with popcorn) on the Wikipedia noticeboard dealing with LLM use: with great regularity, LLM edits have sourcing errors: not ju… | ↗Not briefed |
hn hn-ai | 2mo | Solving Ancient Ciphers with Claude | — | ↗Not briefed |
bluesky bluesky-ai | 2mo | I mean, in that sense I don't think deepseek is a pure distill either. but I absolutely think everyone is covertly collecting model outputs… | I mean, in that sense I don't think deepseek is a pure distill either. but I absolutely think everyone is covertly collecting model ou… | ↗Not briefed |
hn hn-ai | 2mo | From Templates to Reflection, Part 1: Back to the Basics with Type_vector | — | ↗Not briefed |
arxiv arxiv-ai | 2mo | After the Party: Governing What a Viral Agent-Skill Ecosystem Left Behind | AI agents increasingly act through agent skills, i.e., natural-language instructions, that direct a host agent toward shell, network,… | ↗Not briefed |
bluesky bluesky-ai | 2mo | The basic idea is that AI isn’t going to be capable of making important breakthroughs in curing disease unless researchers can feed the mod… | The basic idea is that AI isn’t going to be capable of making important breakthroughs in curing disease unless researchers can feed th… | ↗Not briefed |
bluesky bluesky-ai | 2mo | ‘When developers use Claude Code with a non-Anthropic model, accessed through a proxy or router, Anthropic loses the stream of revenue it w… | ‘When developers use Claude Code with a non-Anthropic model, accessed through a proxy or router, Anthropic loses the stream of revenue… | ↗Not briefed |
bluesky bluesky-ai | 2mo | idk how much of an unpopular opinion it is but in my opinion, allegedly, the openai/anthropic ""whistleblower"" is full of shit and looking… | idk how much of an unpopular opinion it is but in my opinion, allegedly, the openai/anthropic ""whistleblower"" is full of shit and lo… | ↗Not briefed |
hn hn-ai | 2mo | A Cop Searched 19,000 Flock Cameras Across 1,558 Cities. His Reason: 'LMAO' | — | ↗Not briefed |
bluesky bluesky-ai | 2mo | MXFP4 is the format in which deepseek v4.1 flash is released so by definition it's lossless compared to the original release | MXFP4 is the format in which deepseek v4.1 flash is released so by definition it's lossless compared to the original release | ↗Not briefed |
hn hn-ai | 2mo | Online Z3 Guide | — | ↗Not briefed |
bluesky bluesky-ai | 2mo | 7. Everyone would be better off. Well maybe not everyone. Jensen Huang would certainly have a very big sad. But is that a bug or a feature?… | 7. Everyone would be better off. Well maybe not everyone. Jensen Huang would certainly have a very big sad. But is that a bug or a fea… | ↗Not briefed |
bluesky bluesky-ai | 2mo | 6. So their bosses should encourage them to organize, signing cards to build the first labor union willing to go on strike to de-immanentiz… | 6. So their bosses should encourage them to organize, signing cards to build the first labor union willing to go on strike to de-imman… | ↗Not briefed |
bluesky bluesky-ai | 2mo | "A Single Firm is Behind OpenAI, Anthropic, and Meta Hacking Scandals: The Israeli Effective Altruist firm Irregular caused unsecured AI mo… | "A Single Firm is Behind OpenAI, Anthropic, and Meta Hacking Scandals: The Israeli Effective Altruist firm Irregular caused unsecured… | ↗Not briefed |
hn hn-ai | 2mo | Show HN: RelateAnything – Real-time open-vocabulary relation prediction | — | ↗Not briefed |
hn hn-ai | 2mo | An open source alternative to Raycast, in Rust | — | ↗Not briefed |
hn hn-ai | 2mo | MiSeGuard – Deterministic runtime safety layer for AI coding agents | — | ↗Not briefed |
rss techcrunch-ai | 2mo | Former TikTok execs built an app that uses AI to teach you how to pose for a photo | Essentially a camera app, Superpose analyzes selfies or photos and generates four potential poses using AI. | ↗Not briefed |
bluesky bluesky-ai | 2mo | “The 2023 AI Impacts survey of ML researchers put the median at roughly 5% for extremely bad outcomes; superforecaster panels have landed c… | “The 2023 AI Impacts survey of ML researchers put the median at roughly 5% for extremely bad outcomes; superforecaster panels have lan… | ↗Not briefed |
hn hn-ai | 2mo | Into the CUDA Multiverse of Runtime Compilation: GPU DBMS Kernel Fusion [pdf] | — | ↗Not briefed |
arxiv arxiv-ai | 2mo | Towards Illusions Awareness in Cyber-Physical System's Design | Cyber-Physical Systems (CPS) operate through a continuous sense-compute-act loop within an open context environment, making it impossi… | ↗Not briefed |
rss techmeme | 2mo | OpenAI says it has been working with Anthropic and Google for several weeks on AI safety, and it does not need an antitrust waiver to coordinate on safety (Maggie Eastland/Bloomberg Law) | Maggie Eastland / Bloomberg Law: OpenAI says it has been working with Anthropic and Google for several weeks on AI safety, and it does… | ↗Not briefed |
hn hn-ai | 2mo | What we have learned at OpenShell applying formal methods to control AI agents | — | ↗Not briefed |
bluesky bluesky-ai | 2mo | AI has its limits. It is not so good as physical models for predicting weather extremes www.science.org/doi/10.1126/... has little abilit… | AI has its limits. It is not so good as physical models for predicting weather extremes www.science.org/doi/10.1126/... has little a… | ↗Not briefed |
bluesky bluesky-ai | 2mo | 0 Frauen 6 Männer Nen ganzen Podcast produzieren aber ganz ohne Frauen, muss Mann auch erstmal schaffen. Quelle www.instagram.com/p/Dd… | 0 Frauen 6 Männer Nen ganzen Podcast produzieren aber ganz ohne Frauen, muss Mann auch erstmal schaffen. Quelle www.instagram.com… | ↗Not briefed |
bluesky bluesky-ai | 2mo | The AI bot swarm safety alarm is about one thing only. It’s warning every utility and hospital that they need to buy Claude or they are gon… | The AI bot swarm safety alarm is about one thing only. It’s warning every utility and hospital that they need to buy Claude or they ar… | ↗Not briefed |
bluesky bluesky-ai | 2mo | LLMじゃなくて、紙みたいなモニターとか鉛筆みたいに芯を長く削り出せるペンタブレットとか そっちの方に進化してもらいたかった… | LLMじゃなくて、紙みたいなモニターとか鉛筆みたいに芯を長く削り出せるペンタブレットとか そっちの方に進化してもらいたかった… | ↗Not briefed |
bluesky bluesky-ai | 2mo | there are basically two aspects of this story that make it important, in my view: - a couple of guys could ~90% crack a Millennium prize pr… | there are basically two aspects of this story that make it important, in my view: - a couple of guys could ~90% crack a Millennium pri… | ↗Not briefed |
hn hn-ai | 2mo | Ask HN: Why 17? | If I ask an LLM to pick a number between 1 and 30, it always picks 17. Tried with multiple models. Why 17? | ↗Not briefed |
bluesky bluesky-ai | 2mo | Dan: "So, don't write one of those unless you are confident you can control it." Silicon Valley Types, whilst hooking up an executable pro… | Dan: "So, don't write one of those unless you are confident you can control it." Silicon Valley Types, whilst hooking up an executabl… | ↗Not briefed |
arxiv arxiv-ai | 2mo | Persistent Recurrent Memory Between Transformer Layers - Improves Language Model Generalization | We introduce a simple architectural modification to decoder-only transformers: a persistent recurrent state that observes hidden repre… | ↗Not briefed |
hn hn-ai | 2mo | GRP-Obliteration: Unaligning LLMs with a Single Unlabeled Prompt | — | ↗Not briefed |
bluesky bluesky-ai | 2mo | Anthropic said its AI may have been misused for biological weapons development. But these reports only come when AI companies want the pub… | Anthropic said its AI may have been misused for biological weapons development. But these reports only come when AI companies want th… | ↗Not briefed |
bluesky bluesky-ai | 2mo | it's just not the case that LLMs can destroy humanity. An LLM cannot (technical sense) do shit. The thing which can damage important system… | it's just not the case that LLMs can destroy humanity. An LLM cannot (technical sense) do shit. The thing which can damage important s… | ↗Not briefed |
rss techmeme | 2mo | Artificial Intelligence Underwriting Company, which offers audits and safety certifications for AI agents, raised a $40M Series A, bringing total raised to $55M (Julie Bort/TechCrunch) | Julie Bort / TechCrunch: Artificial Intelligence Underwriting Company, which offers audits and safety certifications for AI agents, ra… | ↗Not briefed |
rss techcrunch-ai | 2mo | Discover how to take your startup from prototype to production at TechCrunch Disrupt 2026 | Learn how to scale your startup breakthrough from prototype to production at TechCrunch Disrupt 2026 with scaling leaders, Adrian Macn… | ↗Not briefed |
hn hn-ai | 2mo | Intro to AI Safety, Remastered (2021) [video] | — | ↗Not briefed |
bluesky bluesky-ai | 2mo | I read the original argument and had two thoughts. One, yes repetitiveness is a still problem with basic LLM outputs and sometimes an avail… | I read the original argument and had two thoughts. One, yes repetitiveness is a still problem with basic LLM outputs and sometimes an… | ↗Not briefed |
rss the-decoder | 2mo | After warning AI is too dangerous, Bill Gates bets a billion on its upside | The Gates Foundation is investing at least a billion dollars over two years to make AI tools more widely available in health, educatio… | ↗Not briefed |
bluesky bluesky-ai | 2mo | Tech oligarchs agree calling AI “too dangerous” way cooler than just admitting it doesn’t work 🥰 www.thebeaverton.com/2026/09/tech... #… | Tech oligarchs agree calling AI “too dangerous” way cooler than just admitting it doesn’t work 🥰 www.thebeaverton.com/2026/09/tech.… | ↗Not briefed |
rss techmeme | 2mo | Meta says it plans to deploy MTIA 450, its third-generation custom AI chip, in data centers in H1 2027, followed by a broad rollout of MTIA 500 by end of 2027 (Dina Bass/Bloomberg) | Dina Bass / Bloomberg: Meta says it plans to deploy MTIA 450, its third-generation custom AI chip, in data centers in H1 2027, followe… | ↗Not briefed |
hn hn-ai | 2mo | Show HN: Let agent read files with secrets while redacting values for LLM contex | — | ↗Not briefed |
arxiv arxiv-ai | 2mo | ECHO: Early-layer Collaborative Hierarchical Orchestration with Bonus Logits in Speculative Decoding | While draft-model-free speculative decoding offers a promising path to efficient LLM inference, it is frequently constrained by stale… | ↗Not briefed |
rss techmeme | 2mo | The US DOJ says several companies used Binance to relay payments of around $1.5B for black-market Iranian oil from Chinese buyers; Binance says it took action (Wall Street Journal) | Wall Street Journal: The US DOJ says several companies used Binance to relay payments of around $1.5B for black-market Iranian oil fro… | ↗Not briefed |
bluesky bluesky-ai | 2mo | after a few more days with deepseek v4.1 flash instead of GPT-6 we are now at 395tps prefill and 16tps decode waow | after a few more days with deepseek v4.1 flash instead of GPT-6 we are now at 395tps prefill and 16tps decode waow | ↗Not briefed |
hn hn-ai | 2mo | I Turned Google Glass into a Claude Code Terminal | — | ↗Not briefed |