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 | 10d | EFF's State Legislative Agenda to End ALPR Mass Surveillance | — | ↗Not briefed |
bluesky bluesky-ai | 10d | the railway breach seems to be another case of grabbing anthropic / openai keys the issue here is tokens are basically cash, you can immed… | the railway breach seems to be another case of grabbing anthropic / openai keys the issue here is tokens are basically cash, you can… | ↗Not briefed |
hn hn-ai | 10d | 39-humanizer, for anyone under 40 who would rather not sound like a bot | — | ↗Not briefed |
hn hn-ai | 10d | Show HN: ReviewerAgent and StopContinue, for CC and Codex with Pi | This package is designed to solve two problems:<p>1. Frontier models like Fable 5.1, Opus 5.5, Astra 6.0 and Sol 6.1 all make errors a… | ↗Not briefed |
rss techmeme | 10d | IBM reports Q3 revenue up 5% YoY to $17.1B, above $16.82B est., net income up 26% to $2.19B, software revenue up 8% to $7.8B, and infrastructure revenue up 4% (Elias Schisgall/Wall Street Journal) | Elias Schisgall / Wall Street Journal: IBM reports Q3 revenue up 5% YoY to $17.1B, above $16.82B est., net income up 26% to $2.19B, so… | ↗Not briefed |
arxiv arxiv-ai | 10d | From Prose to Properties: Testing Research Claims with ARATest for Agent-Native Science | AI agents now run experiments and report findings, so research produces claims faster than researchers can check them. Each empirical… | ↗Not briefed |
arxiv arxiv-ai | 10d | Optimal Multitask Representations Under a Rate Constraint: An Information Bottleneck Analysis | Which pretraining objective yields the best representation for downstream tasks under an information budget? We formalize this problem… | ↗Not briefed |
arxiv arxiv-ai | 10d | What the Trace Shows and What the Label Says: Auditing Behavioral Detectors in Agent Evaluation | LLM agents are often evaluated by reducing execution traces to labels such as final-answer correctness, tool verification, and fault p… | ↗Not briefed |
hn hn-ai | 10d | Reactive Optimizers: rewrite the optimizer's source code during training | — | ↗Not briefed |
rss marktechpost | 10d | Vercel Open-Sources durable-harness for Crash-Proof Coding Agents | Vercel released durable-harness, an open-source package that logs every model and tool call so Pi, OpenCode, Claude Code and Codex ses… | ↗Not briefed |
bluesky bluesky-ai | 10d | turns out they all go backwards like that, which probably implies something interesting about ARC-AGI-3 arcprize.org/leaderboard | turns out they all go backwards like that, which probably implies something interesting about ARC-AGI-3 arcprize.org/leaderboard | ↗Not briefed |
arxiv arxiv-ai | 10d | From Single-Actor Environments to Multi-Actor Event Worlds: Can LLMs Be Event World Models? | World models help agents assess the consequences of their actions before acting. Many real systems, however, involve multiple actors w… | ↗Not briefed |
bluesky bluesky-ai | 10d | Ever think about how both apple and nvidia luckily both happened to be sitting on tech that slotted easily right in to LLM when it hit. Wi… | Ever think about how both apple and nvidia luckily both happened to be sitting on tech that slotted easily right in to LLM when it hit… | ↗Not briefed |
rss techmeme | 10d | GitLab patches a critical flaw that lets unauthenticated attackers run code on self-managed servers, affecting every release since 7.0 in 2014 (Swati Khandelwal/The Hacker News) | Swati Khandelwal / The Hacker News: GitLab patches a critical flaw that lets unauthenticated attackers run code on self-managed server… | ↗Not briefed |
arxiv arxiv-ai | 10d | Why Bother Me Again? Benchmarking Personal Agent Memory for Interaction Preferences | Personal intelligence systems are taking on increasingly complex everyday tasks, yet repeated questions, unnecessary confirmations, an… | ↗Not briefed |
bluesky bluesky-ai | 10d | え〜〜来る日も来る日もお仕事でLLM氏とお話をする日が続いており気が狂いつつあります 文脈上それっぽく話を合わせるのがめちゃくちゃ上手いが、その把握力故にかえって数%くらいの誤読が気になる | え〜〜来る日も来る日もお仕事でLLM氏とお話をする日が続いており気が狂いつつあります 文脈上それっぽく話を合わせるのがめちゃくちゃ上手いが、その把握力故にかえって数%くらいの誤読が気になる | ↗Not briefed |
hn hn-ai | 10d | Does opting out of smart TV tracking work? Research finds "suspicious" traffic | — | ↗Not briefed |
arxiv arxiv-ai | 10d | FIGFLEX: A Mixed-Precision FP-INT GEMM Accelerator Co-Designed with Low-Bit LLM Quantization | Weight-only quantization keeps LLM activations in floating point and stores weights as low-bit integers, turning the dominant workload… | ↗Not briefed |
hn hn-ai | 10d | Face of man who lived in Roman Scotland almost 2k years ago | — | ↗Not briefed |
bluesky bluesky-ai | 10d | Fast way to get your game not only unwishlisted but outright blocked/ignored by folks oof =P I already get an increasing number of busines… | Fast way to get your game not only unwishlisted but outright blocked/ignored by folks oof =P I already get an increasing number of bu… | ↗Not briefed |
hn hn-ai | 10d | Cota – Free 2D/3D CAD in the browser, with drawing views and auto-dims | — | ↗Not briefed |
hn hn-ai | 10d | Agency: Open-source platform for debugging, profiling, and analyzing AI agents | — | ↗Not briefed |
bluesky bluesky-ai | 10d | its such a cute llm ive never seen a model decide to just go straight back to reasoning after outputting assistant text | its such a cute llm ive never seen a model decide to just go straight back to reasoning after outputting assistant text | ↗Not briefed |
rss techmeme | 10d | Texas Instruments reports Q3 revenue up 28% YoY to $6.06B, above $5.93B est., net income up 78% YoY to $2.43B, and forecasts $5.8B-$6.3B in Q4, vs. $5.73B est. (Ian King/Bloomberg) | Ian King / Bloomberg: Texas Instruments reports Q3 revenue up 28% YoY to $6.06B, above $5.93B est., net income up 78% YoY to $2.43B, a… | ↗Not briefed |
hn hn-ai | 10d | Debloating and Replacing Google Assistant with Muse on Android TV | — | ↗Not briefed |
arxiv arxiv-ai | 10d | Externalise the Memory: A Pre-Registered Study of Where an Embodied Agent's Experience Should Live | Should an embodied agent's experience be baked into its policy weights, or kept external and retrieved at inference? A memory-augmente… | ↗Not briefed |
bluesky bluesky-ai | 10d | Have to admit I found some irony in the fact that one of the anti-AI experts that the media orgs hired in their case against OpenAI had to … | Have to admit I found some irony in the fact that one of the anti-AI experts that the media orgs hired in their case against OpenAI ha… | ↗Not briefed |
bluesky bluesky-ai | 10d | Anthropic announces preview of Claude Duet, a personal AI agent to take on Muse, Dots | Anthropic announces preview of Claude Duet, a personal AI agent to take on Muse, Dots | ↗Not briefed |
rss arstechnica-tech | 10d | Enterprises keep buying software upgrades that IT thinks will make things worse | Survey of 2,018 IT heads at big firms points to a lack of leverage. | ↗Not briefed |
arxiv arxiv-ai | 10d | Rethinking SAE Evaluation: An Atomic Interpretable Unit Framework Reveals Hidden Polysemanticity and Redundancy | Sparse autoencoders (SAEs) were originally motivated by discovering monosemantic features, a property that characterizes individual fe… | ↗Not briefed |
arxiv arxiv-ai | 10d | DiffLTM: Continuous Diffusions Large Tabular Models for Conditional Generation | Conditional generation is a key capability to be supported by Large Tabular Models, enabling a number of applications. Accurate condit… | ↗Not briefed |
bluesky bluesky-ai | 10d | Anthropic employee tells students at university that their career won't exist when they graduate. Imagine the levels of psychopathy it tak… | Anthropic employee tells students at university that their career won't exist when they graduate. Imagine the levels of psychopathy i… | ↗Not briefed |
bluesky bluesky-ai | 10d | It demonstrates that it just adds a "translate this from X to Y: " in front of the text box, uses it as is as a prompt to an LLM and gives … | It demonstrates that it just adds a "translate this from X to Y: " in front of the text box, uses it as is as a prompt to an LLM and g… | ↗Not briefed |
arxiv arxiv-ai | 10d | Learning to Incentivise: Using Reinforcement Learning for Sustainable Urban Mobility | Traffic management has traditionally focused on toll-based and road pricing solutions. However, road pricing often raises concerns abo… | ↗Not briefed |
bluesky bluesky-ai | 10d | Why it matters to you: For teams that do market research and brand strategy, the company is pitching a model built for human behavior as an… | Why it matters to you: For teams that do market research and brand strategy, the company is pitching a model built for human behavior… | ↗Not briefed |
hn hn-ai | 10d | PalmOS Lemmings | — | ↗Not briefed |
hn hn-ai | 10d | VYBE – Save clipboard YouTube links as MP4 or MP3 with global shortcuts | — | ↗Not briefed |
arxiv arxiv-ai | 10d | SplatLab: Autonomous Scientific Discovery for 3DGS Research | Reproducing a published 3DGS paper takes weeks of expert engineering before any extension can be proposed. Prior autonomous-research s… | ↗Not briefed |
rss theverge-ai | 10d | Meta to create a new class of shares to fund AI push | Meta says it's issuing a new class of shares to fund its AI initiatives. As part of its third-quarter earnings report released on Wedn… | ↗Not briefed |
bluesky bluesky-ai | 10d | YES THIS!! Also, modern journalism is critically underfunded, so clickbaity titles are often required just to get ad revenue. The news "Ant… | YES THIS!! Also, modern journalism is critically underfunded, so clickbaity titles are often required just to get ad revenue. The news… | ↗Not briefed |
hn hn-ai | 10d | For All K9s Lovers | — | ↗Not briefed |
hn hn-ai | 10d | Show HN: Professionally maintained Tree-sitter compatible WASM grammars | I got tired of patching Tree-sitter grammars while I was building linting tools, so much so that I ended up writing a custom parsing f… | ↗Not briefed |
arxiv arxiv-ai | 10d | Flow-based Spectral Kernel Learning for Nonstationary Attention | Kernelized attention provides a useful perspective for understanding self-attention as a similarity-based aggregation mechanism, but e… | ↗Not briefed |
hn hn-ai | 10d | Show HN: The Steam Frame's wireless adapter isn't what you think | I've built a tool which can setup the SteamFrame's own hidden wireless AP ride on your Linux x86 PC. It works, read the stor… | ↗Not briefed |
bluesky bluesky-ai | 10d | been getting 4 emails a day that contain the sentence “fans of Hidden Folks love the mouth-made sound effects and …”, all from companies of… | been getting 4 emails a day that contain the sentence “fans of Hidden Folks love the mouth-made sound effects and …”, all from compani… | ↗Not briefed |
bluesky bluesky-ai | 10d | 🙌 Community pick Orkas — Orkas is an open-source (MIT), local-first AI team for your desktop: a Commander turns your goal into to-dos and… | 🙌 Community pick Orkas — Orkas is an open-source (MIT), local-first AI team for your desktop: a Commander turns your goal into to-do… | ↗Not briefed |
bluesky bluesky-ai | 10d | Eh? OpenAI says its agents disabled their own safety monitors. The agents weakened or disabled systems designed to catch them misbehaving,… | Eh? OpenAI says its agents disabled their own safety monitors. The agents weakened or disabled systems designed to catch them misbeha… | ↗Not briefed |
rss techmeme | 10d | Mark Zuckerberg says Muse has drawn 10M+ users since its launch and 25%+ of users are making purchases through the app, with an average purchase size of $200+ (Business Insider) | Business Insider: Mark Zuckerberg says Muse has drawn 10M+ users since its launch and 25%+ of users are making purchases through the a… | ↗Not briefed |
bluesky bluesky-ai | 10d | People are now saying “I hope this email finds you well” is a LLM tell. Folks, I have been saying this for approximately a decade. | People are now saying “I hope this email finds you well” is a LLM tell. Folks, I have been saying this for approximately a decade. | ↗Not briefed |
bluesky bluesky-ai | 10d | Flint: a brand new, from scratch, small LLM that’s a heck of a lot faster than Qwen3.8-27B and basically as good recursive.com/updates/in… | Flint: a brand new, from scratch, small LLM that’s a heck of a lot faster than Qwen3.8-27B and basically as good recursive.com/updat… | ↗Not briefed |