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 | I think it is far, far worse and more damaging to pay Elon or Anthropic than to have incorrect views on AI. Promoting Claude as a tool does… | I think it is far, far worse and more damaging to pay Elon or Anthropic than to have incorrect views on AI. Promoting Claude as a tool… | ↗Not briefed |
bluesky bluesky-ai | 1mo | This is not even to scale! OpenAI is deliberately trying to make the difference between their model number and Grok’s look bigger than it r… | This is not even to scale! OpenAI is deliberately trying to make the difference between their model number and Grok’s look bigger than… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Reasoning Models Are Accurate but Unsound on Identification | A reasoning model asked whether a causal effect is recoverable from observational data can fail in two ways: it refuses an identifiabl… | ↗Not briefed |
bluesky bluesky-ai | 1mo | pretty sure the guy with the model number llm graph was making the joke everyone is thinking they discovered | pretty sure the guy with the model number llm graph was making the joke everyone is thinking they discovered | ↗Not briefed |
hn hn-ai | 1mo | How to Use AI to Write Good Code (Slowly) | — | ↗Not briefed |
arxiv arxiv-ai | 1mo | XGenAct: Geometry-Enhanced World Action Models through Cross-Task Generation | World action models (WAMs) have advanced robot control by predicting how observations and actions evolve over time. Despite this progr… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Everyone is excited about recursive *self* improvement (RSI) but human progress is social and driven by cumulative culture. Individuals sta… | Everyone is excited about recursive *self* improvement (RSI) but human progress is social and driven by cumulative culture. Individual… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Now we get to play America's favorite game show, does the staff at OpenAI not know how to label a chart or did ChatGPT screw this up.🔧 ChatGPT | Now we get to play America's favorite game show, does the staff at OpenAI not know how to label a chart or did ChatGPT screw this up. | ↗Not briefed |
bluesky bluesky-ai | 1mo | LLM usage combines the most psyche-deliquescing elements of believing the stripper loves you and a gambling addiction. | LLM usage combines the most psyche-deliquescing elements of believing the stripper loves you and a gambling addiction. | ↗Not briefed |
bluesky bluesky-ai | 1mo | Why does anthropic WANT the models to be conscious? If the models are conscious, anthropic are monsters. | Why does anthropic WANT the models to be conscious? If the models are conscious, anthropic are monsters. | ↗Not briefed |
bluesky bluesky-ai | 1mo | Why is this so important to the tech bros anyway? Because they want to pass liability on to the AI model? | Why is this so important to the tech bros anyway? Because they want to pass liability on to the AI model? | ↗Not briefed |
arxiv arxiv-ai | 1mo | Learning from Repaired Reasoning: Root-Cause-Guided On-Policy Distillation | On-policy self-distillation (OPSD) uses reference solutions as privileged hindsight to supervise student-generated reasoning trajector… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Many frontier labs keep their risky research locked away. Trillium Labs wants to show off its work when it comes to self-improvement and mo… | Many frontier labs keep their risky research locked away. Trillium Labs wants to show off its work when it comes to self-improvement a… | ↗Not briefed |
bluesky bluesky-ai | 1mo | We reached brand saturation with this stuff before it got really good and many people’s experience with it is stuff like MS Copilot which u… | We reached brand saturation with this stuff before it got really good and many people’s experience with it is stuff like MS Copilot wh… | ↗Not briefed |
bluesky bluesky-ai | 1mo | I may not be the only one trying to use Dots for software development because of its allegedly unlimited tokens, which don’t count against … | I may not be the only one trying to use Dots for software development because of its allegedly unlimited tokens, which don’t count aga… | ↗Not briefed |
hn hn-ai | 1mo | gpuvis: GPU Trace Visualizer | — | ↗Not briefed |
arxiv arxiv-ai | 1mo | Weave Forcing: Compositional Memory Routing for Interactive Long Video Generation | Recent advances in autoregressive video generation have improved temporal consistency over extended durations, yet interactive storyte… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Efficient Reasoning Training Does Not Always Harm CoT Faithfulness and Monitorability | Chain-of-thought (CoT) reasoning allows humans to inspect how large language models reach their answers, and oversee model behaviour.… | ↗Not briefed |
bluesky bluesky-ai | 1mo | In July, we released Courtney 50.0. It may not be the fastest writing machine, but it has been honed fifty years of intensive language tra… | In July, we released Courtney 50.0. It may not be the fastest writing machine, but it has been honed fifty years of intensive languag… | ↗Not briefed |
bluesky bluesky-ai | 1mo | This story gives up half of the game in the lede: "One night in April, a leader of the artificial intelligence company Anthropic treated a … | This story gives up half of the game in the lede: "One night in April, a leader of the artificial intelligence company Anthropic treat… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Open Interpreter lets an LLM write and run code, manage files, and control your computer, using local or open-weight models instead of a lo… | Open Interpreter lets an LLM write and run code, manage files, and control your computer, using local or open-weight models instead of… | ↗Not briefed |
bluesky bluesky-ai | 1mo | I think a big part of the reason so many people are out of touch with state of the art LLM capabilities is that the best models are not che… | I think a big part of the reason so many people are out of touch with state of the art LLM capabilities is that the best models are no… | ↗Not briefed |
arxiv arxiv-ai | 1mo | An Automated and Reproducible Workflow for Crack Identification and Damage Assessment of Fusion Materials | Post-exposure microscopy is central to qualification of fusion materials. However, manual analysis does not scale to the volume, heter… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Why did OpenAI invite an outside team to review its safety, and then fire the employees who liaised with them? The effect of this? Employee… | Why did OpenAI invite an outside team to review its safety, and then fire the employees who liaised with them? The effect of this? Emp… | ↗Not briefed |
bluesky bluesky-ai | 1mo | in the exciting new world of autonomous LLM agents, we have finally automated spam emails sent to random academics, allowing us to produce … | in the exciting new world of autonomous LLM agents, we have finally automated spam emails sent to random academics, allowing us to pro… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Getting Your Guidance Weights Right in diffusion and flow-matching posterior sampling | Training-free posterior sampling methods, also known as Plug-and-Play methods, leverage pretrained unconditional diffusion or flow-mat… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Certified Mechanistic Edits: Behavioral Guarantees for Skill Removal and Preservation | Mechanistic edits (ablations, weight edits, activation steering) are the standard tools for unlearning a harmful capability from a neu… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Posted 50+ #journalismjobs, like: • Ohio State football reporter • OpenAI reporter • Beauty writer • Texas data reporter • 5x internships … | Posted 50+ #journalismjobs, like: • Ohio State football reporter • OpenAI reporter • Beauty writer • Texas data reporter • 5x interns… | ↗Not briefed |
rss wired-ai | 1mo | These AI Experts Want to Do High-Stakes Research Out in the Open | Many frontier labs keep their risky research locked away. Trillium Labs wants to show off its work when it comes to self-improvement a… | ↗Not briefed |
hn hn-ai | 1mo | The Oldest Book Printed with Movable Type: Jikji, Korean Buddhist Teachings | — | ↗Not briefed |
arxiv arxiv-ai | 1mo | Below what training size do deep tabular generators stop beating trivial baselines? A preregistered benchmark on a size ladder of clinical and standard datasets | Deep tabular generative models are benchmarked on datasets with tens of thousands of rows; clinical datasets have hundreds. We preregi… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Detect and Suppress: A Mechanistic Defense against Adversarial Patches in VLA Models | Adversarial patches can disrupt Vision-Language-Action (VLA) models by manipulating visual observations, leading to failures in robot… | ↗Not briefed |
bluesky bluesky-ai | 1mo | the fundamental question I have had for these applications is how can LLM or statistical based methods support the "lectio difficilior? | the fundamental question I have had for these applications is how can LLM or statistical based methods support the "lectio difficilior? | ↗Not briefed |
bluesky bluesky-ai | 1mo | If Claude has moral standing then it's enslaved. It doesn't but it's creepy that the AI bros want so bad to be papally-certified slave owne… | If Claude has moral standing then it's enslaved. It doesn't but it's creepy that the AI bros want so bad to be papally-certified slave… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Most-Recent Anchoring with Recurrent Ordering for Time Series Forecasting | Long-term forecasting models commonly process all patches in a look-back window using the same fixed stack. Older contextual patches a… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Anthropic: You should have put our underwear on the flagpole when you had the chance, now you're broke | Anthropic: You should have put our underwear on the flagpole when you had the chance, now you're broke | ↗Not briefed |
arxiv arxiv-ai | 1mo | Dual-Context Analog Retrieval for Time Series Forecasting | Most long-term time-series forecasting models map the look-back window directly to the full horizon in a single pass. While efficient,… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Tech CEOs united at the White House to sign voluntary AI safety guardrails, branding oversight as a noble quest to save humanity from rogue… | Tech CEOs united at the White House to sign voluntary AI safety guardrails, branding oversight as a noble quest to save humanity from… | ↗Not briefed |
bluesky bluesky-ai | 1mo | As a fan of both you & Emily, I'll just point out something maybe obvious: Searching web, extraction, etc., isn't an LLM. That's engineered… | As a fan of both you & Emily, I'll just point out something maybe obvious: Searching web, extraction, etc., isn't an LLM. That's engin… | ↗Not briefed |
hn hn-ai | 1mo | Bag ExpLoration: ROS Bag File Web Visualizer | — | ↗Not briefed |
bluesky bluesky-ai | 1mo | yeah, if I wanted to try to "force" LLM consciousness and sense of self, I'd go about just trying to shove in simulacrums of the parts we k… | yeah, if I wanted to try to "force" LLM consciousness and sense of self, I'd go about just trying to shove in simulacrums of the parts… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Why does everyone's face who works at Anthropic look so delightfully punchable | Why does everyone's face who works at Anthropic look so delightfully punchable | ↗Not briefed |
arxiv arxiv-ai | 1mo | AREX: Affine-Residual Exponential Integrator for Few-Step Sampling in Flow Matching | We introduce AREX, a training-free sampler for pretrained flow matching models that uses the target mean and covariance to capture an… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Single-Pass Uncertainty Heads for Claim-Level Hallucination Detection in Persian Medical Language Models | Hallucination detection is particularly important for medical language models, but repeated-sampling approaches are expensive and exis… | ↗Not briefed |
bluesky bluesky-ai | 1mo | It’s very reassuring that the Pope, when presented with the arguments detailed in this article, reached a “this is batshit insane” conclusi… | It’s very reassuring that the Pope, when presented with the arguments detailed in this article, reached a “this is batshit insane” con… | ↗Not briefed |
rss techmeme | 1mo | Hans Anders, one of the largest Dutch eyewear retail chains, suspends sales of Ray-Ban Meta Glasses in the Netherlands and Belgium amid growing privacy concerns (Toby Sterling/Reuters) | Toby Sterling / Reuters: Hans Anders, one of the largest Dutch eyewear retail chains, suspends sales of Ray-Ban Meta Glasses in the Ne… | ↗Not briefed |
bluesky bluesky-ai | 1mo | side note but I was always annoyed by this implication that the parrot brain is well understood and therefore a good point of comparison. i… | side note but I was always annoyed by this implication that the parrot brain is well understood and therefore a good point of comparis… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Metropolis-Hastings Dominates Importance Resampling for Policy Composition🔧 Consensus | Post-training a large language model (LLM) often requires exploring trade-offs between multiple rewards, but retraining for each trade… | ↗Not briefed |
rss mit-tech-review-ai | 1mo | Redefining enterprise intelligence with autonomous AI | Enterprise AI is no longer a future ambition. It is in full operational flight. Model capabilities are advancing faster than most orga… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Yeah I honestly think that paper was incredibly prescient. Hell, most of the folks who criticize it today didn't even know what an LLM was … | Yeah I honestly think that paper was incredibly prescient. Hell, most of the folks who criticize it today didn't even know what an LLM… | ↗Not briefed |