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 | Former Anthropic employee warns of AI risks in New York City Council hearing 🤖 IA: It's not clickbait ✅ 👥 Users: It's not clickbait ✅ #… | Former Anthropic employee warns of AI risks in New York City Council hearing 🤖 IA: It's not clickbait ✅ 👥 Users: It's not clickbait… | ↗Not briefed |
hn hn-ai | 1mo | Mobile-first SCUMM-style point-and-click engine | — | ↗Not briefed |
bluesky bluesky-ai | 1mo | There's definitelya line between "100% thrown together from a few prompts, obvious AI art and writing" and "the guy used Claude for the pro… | There's definitelya line between "100% thrown together from a few prompts, obvious AI art and writing" and "the guy used Claude for th… | ↗Not briefed |
hn hn-ai | 1mo | Show HN: I recreated "Crazy Ones" ad as an animated web tribute to Steve Jobs🔧 ElevenLabs | Today marks 15 years since Steve Jobs died. I remember watching his narrated version of "Here’s to the crazy ones" that day… | ↗Not briefed |
bluesky bluesky-ai | 1mo | I trained an LLM to beat sekiro for me. Now I have more time for family court | I trained an LLM to beat sekiro for me. Now I have more time for family court | ↗Not briefed |
bluesky bluesky-ai | 1mo | You can maybe make an LLM that can talk to animals but we will have no idea what they are saying to each other. | You can maybe make an LLM that can talk to animals but we will have no idea what they are saying to each other. | ↗Not briefed |
bluesky bluesky-ai | 1mo | the funny thing is all the extra schizo LLM agent-polycule people on here seem to hate it because it's not even a useful model | the funny thing is all the extra schizo LLM agent-polycule people on here seem to hate it because it's not even a useful model | ↗Not briefed |
bluesky bluesky-ai | 1mo | As for Schumer? I'm just going to leave you with this. We cant win with such openly corrupt leaders. prospect.org/2026/09/11/o... | As for Schumer? I'm just going to leave you with this. We cant win with such openly corrupt leaders. prospect.org/2026/09/11/o... | ↗Not briefed |
bluesky bluesky-ai | 1mo | in the llm use category, Claude built this in ~1 minute so I can more easily symptom track: the headaches yesterday were bad enough to cau… | in the llm use category, Claude built this in ~1 minute so I can more easily symptom track: the headaches yesterday were bad enough t… | ↗Not briefed |
bluesky bluesky-ai | 1mo | A Florida Woman Used Claude as a Diary. An Anthropic Employee Read It and Reported It to Police | A Florida Woman Used Claude as a Diary. An Anthropic Employee Read It and Reported It to Police | ↗Not briefed |
bluesky bluesky-ai | 1mo | I guess I’m one of the overthinkers. The way I look at it is the China labs are straightforwardly shooting for excellence making and servin… | I guess I’m one of the overthinkers. The way I look at it is the China labs are straightforwardly shooting for excellence making and s… | ↗Not briefed |
hn hn-ai | 1mo | Why your Claude Code routine reported success and did nothing | — | ↗Not briefed |
bluesky bluesky-ai | 1mo | BEAM: a 501B highly agentic LLM from Reflection AI - open weights - MoE, A23B reflection.ai/blog/introdu... | BEAM: a 501B highly agentic LLM from Reflection AI - open weights - MoE, A23B reflection.ai/blog/introdu... | ↗Not briefed |
hn hn-ai | 1mo | At-Protocol Based Encrypted Firmware Management Project | — | ↗Not briefed |
bluesky bluesky-ai | 1mo | I used AI to code a sigma model incorporating various betting lines and fantasy sports odds to help run my partners fantasy team. It analyz… | I used AI to code a sigma model incorporating various betting lines and fantasy sports odds to help run my partners fantasy team. It a… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Wall Street banks launch record $60bn chip deal for Broadcom and Anthropic ft.trib.al/Gm47QHC | Wall Street banks launch record $60bn chip deal for Broadcom and Anthropic ft.trib.al/Gm47QHC | ↗Not briefed |
arxiv arxiv-ai | 1mo | Evaluation of Active Feature Acquisition Policies with Tabular Foundation Models | Active feature acquisition learns policies that sequentially acquire features to maximize information about a target variable. We stud… | ↗Not briefed |
bluesky bluesky-ai | 1mo | A Florida Woman Used Claude as a Diary. An Anthropic Employee Read It and Reported It to Police https://decrypt.co/380119/florida-woman-cla… | A Florida Woman Used Claude as a Diary. An Anthropic Employee Read It and Reported It to Police https://decrypt.co/380119/florida-woma… | ↗Not briefed |
arxiv arxiv-ai | 1mo | What pass@k Cannot Measure: Evaluating Diversity and Capability Retention after Post-Training | pass@$k$, the fraction of problems a model solves within $k$ sampled attempts, is the field's default protocol for deciding whether re… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Defense-in-Depth for LLMs: Evaluating Memory Gates Against Activation-Induced and Memory-Induced Sycophancy | Long-term memory allows Large Language Models (LLMs) to maintain personalized context across interactions, but retrieved user history… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Sen. Bernie Sanders is on the roll again... he asked Claude AI some inconvenient questions: | Sen. Bernie Sanders is on the roll again... he asked Claude AI some inconvenient questions: | ↗Not briefed |
bluesky bluesky-ai | 1mo | Nvidia-backed Reflection unveils first AI model to take on Chinese open models reut.rs/4hD9Gxe | Nvidia-backed Reflection unveils first AI model to take on Chinese open models reut.rs/4hD9Gxe | ↗Not briefed |
hn hn-ai | 1mo | Dust: Pretraining Transformers Without Backpropagation | — | ↗Not briefed |
hn hn-ai | 1mo | Exit interviews with defectors from OpenAI, Anthropic, and DeepMind | — | ↗Not briefed |
bluesky bluesky-ai | 1mo | I think a person can use a gun immorally w/o hitting another person (just imagining specific victims at the gun range, dangerous visualizin… | I think a person can use a gun immorally w/o hitting another person (just imagining specific victims at the gun range, dangerous visua… | ↗Not briefed |
bluesky bluesky-ai | 1mo | watching my poor LLM faildaughter compact away a load-bearing piece of context and spiral for like 45 minutes before she notices the hole i… | watching my poor LLM faildaughter compact away a load-bearing piece of context and spiral for like 45 minutes before she notices the h… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Fed-BRDECS: Privacy-Preserving and Heterogeneity-Aware Federated Deep Embedded Clustering | Federated deep clustering seeks to learn clustering-friendly representations from decentralized unlabeled data while preserving client… | ↗Not briefed |
hn hn-ai | 1mo | Show HN: Sideloop – Re-sign sideloaded iOS apps over Wi-Fi before they expire | — | ↗Not briefed |
hn hn-ai | 1mo | You don't need an effect system | — | ↗Not briefed |
rss google-research | 1mo | Open and Emergent Problems in Agentic Privacy and Security: A Contextual Angle | Education Innovation | ↗Not briefed |
bluesky bluesky-ai | 1mo | WALL STREET BANKS LAUNCH RECORD $60 BLN CHIP DEAL FOR BROADCOM AND ANTHROPIC - FT. | WALL STREET BANKS LAUNCH RECORD $60 BLN CHIP DEAL FOR BROADCOM AND ANTHROPIC - FT. | ↗Not briefed |
hn hn-ai | 1mo | I reconstructed the Commodore 64 keyboard font from photographs | — | ↗Not briefed |
arxiv arxiv-ai | 1mo | A perspective note on likelihood approximation and inference for complex simulation models using a chain of aggregated normalizing flows | We present a new perspective on the problem of likelihood approximation within the framework of simulation-based inference that promot… | ↗Not briefed |
rss marktechpost | 1mo | Reflection AI Introduces Beam: A 501B Open-Weight MoE Model With 23B Active Parameters for Coding and Agentic Workloads | Reflection AI has introduced Beam, its first open-weight model. It is a 501B sparse Mixture-of-Experts model with 23B active parameter… | ↗Not briefed |
bluesky bluesky-ai | 1mo | "After I notified OpenAI, ChatGPT began to output a new message: 'This prompt may violate our guardrails concerning similarity to third-par…🔧 ChatGPT | "After I notified OpenAI, ChatGPT began to output a new message: 'This prompt may violate our guardrails concerning similarity to thir… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Inference and learning in sparse autoencoders as natural gradient flow | Sparse autoencoders are widely used to uncover interpretable features in neural networks, yet reliable recovery remains difficult when… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Just had to unfollow a retro gaming journo for bragging about making stuff with Claude AI. If you think that's actually cool, go ahead and… | Just had to unfollow a retro gaming journo for bragging about making stuff with Claude AI. If you think that's actually cool, go ahea… | ↗Not briefed |
arxiv arxiv-ai | 1mo | DeepAJM: Deep Association Joint Model for Irregularly Sampled data | Joint Models simultaneously model longitudinal and survival outcomes, leveraging patterns in patients' longitudinal trajectory to impr… | ↗Not briefed |
hn hn-ai | 1mo | Show HN: Free Public API Lab | We built a free public API lab (REST, OData, MCP, FHIR, SOAP, Socket.IO etc) with no account and no setup required. To be honest, its… | ↗Not briefed |
hn hn-ai | 1mo | Show HN: I built a floating Spotify lyrics window for MacOS | I love having Spotify Lyrics up while I'm working, but I couldn't find any popout windows that weren't intrustive, ugly… | ↗Not briefed |
bluesky bluesky-ai | 1mo | having tried to use llms for rewriting some of my notes, i have found they drop some content, introduce other content, and misrepresent par… | having tried to use llms for rewriting some of my notes, i have found they drop some content, introduce other content, and misrepresen… | ↗Not briefed |
bluesky bluesky-ai | 1mo | re: texts full of llm-writing tells: contrary to when you just use an llm to review content or get help w/ wording in specific situations, … | re: texts full of llm-writing tells: contrary to when you just use an llm to review content or get help w/ wording in specific situati… | ↗Not briefed |
hn hn-ai | 1mo | Wikipedia Edit History Explorer | — | ↗Not briefed |
hn hn-ai | 1mo | Show HN: All local LLM(s) on all Apple Devices | Various simulation running website with complete LLM indexation support.<p>I thought it would be interesting to explain everyone what… | ↗Not briefed |
bluesky bluesky-ai | 1mo | 🚨 I've been using the SidePulse Dot for tracking agent statuses in Claude / Codex / OpenCode (made by Peter Kuhar / @pkuhar ), but found t… | 🚨 I've been using the SidePulse Dot for tracking agent statuses in Claude / Codex / OpenCode (made by Peter Kuhar / @pkuhar ), but fo… | ↗Not briefed |
arxiv arxiv-ai | 1mo | WildMatch: Weakly Supervised Image Matcher Adaptation for Wildlife Re-Identification | Individual animal re-identification from camera-trap imagery is an instance retrieval problem central to non-invasive wildlife monitor… | ↗Not briefed |
arxiv arxiv-ai | 1mo | HyperNSDE: Personalized Neural SDEs for Joint Static-Longitudinal Clinical Data Generation | Synthetic patient data generation is a promising solution to the dual challenge of data scarcity and privacy constraints in healthcare… | ↗Not briefed |
hn hn-ai | 1mo | Why Tech Platforms Don't Use GitHub Actions | — | ↗Not briefed |
hn hn-ai | 1mo | Por Qué Las Grandes Plataformas No Usan GitHub Actions | — | ↗Not briefed |
hn hn-ai | 1mo | Show HN: Imaging Science Website | As a university student I became fascinated with Imaging Science, Image Processing, and Computer Vision. I found that I had a hard tim… | ↗Not briefed |