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 | Reconstruct, Practice, Go Real: Guided Self-Improvement for Embodied Agents | Building reliable robot capabilities across diverse tasks requires substantial human effort to develop and maintain skills, design rew… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Embedding Prediction Helps Image Generation | In diffusion transformers, a class label or a text prompt is embedded once, and the same condition is reused at every denoising step.… | ↗Not briefed |
arxiv arxiv-ai | 1mo | ScholarCatalyst: A Benchmark for Retrieving Papers That Inspire New Research | What makes great scientists great? Even as AI systems start to make progress on open problems, scientists remain far ahead of them at… | ↗Not briefed |
arxiv arxiv-ai | 1mo | SILSA: Sliding-Window Slice Latents for Topology-Preserving High-Resolution 3D Generation | High-resolution 3D generation increasingly relies on voxel latents and multi-stage pipelines that first predict active structure and t… | ↗Not briefed |
arxiv arxiv-ai | 1mo | VISTA: A Visual Harness for Reasoning in an Interactive World | We show that multimodal models possess strong reasoning abilities and that an appropriate harness can unlock their potential to solve… | ↗Not briefed |
arxiv arxiv-ai | 1mo | TACO: Ternary Absolute-max Column-wise One-sparse Optimizer for LLM Fine-Tuning | Full-parameter fine-tuning of large language models (LLMs) incurs substantial optimizer state memory overhead, limiting the model size… | ↗Not briefed |
hn hn-ai | 1mo | Show HN: The Zoomist – looking and listening to ordinary things more closely | I’ve been building a small project called The Zoomist.<p>The idea is to take ordinary things that we normally see or hear without thin… | ↗Not briefed |
arxiv arxiv-ai | 1mo | FERPO: Forward Entropy-Regularized Policy Optimization | Several state-of-the-art methods for online reinforcement learning in continuous control improve policies using action gradients of a… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Hierarchical Continuous Diffusion Language Models | Discrete diffusion language models offer a compelling alternative to autoregressive generation for tasks demanding bidirectional reaso… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Cost-augmented Schrödinger bridges on graphs are exactly solvable: a Feynman-Kac tilt replaces learned control | The generalized Schrödinger bridge on a graph moves mass between two distributions while charging a cost for the states visited. It ha… | ↗Not briefed |
arxiv arxiv-ai | 1mo | The Missing Primitive: Diagnosing and Repairing Mathematical Reasoning in Large Language Models | While Large Language Models (LLMs) have demonstrated striking capabilities on frontier mathematical problems, it remains unclear wheth… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Trust the Direction, Search the Step: Zero-and-First-Order Methods for LLM Fine-Tuning | Step-size selection remains a central challenge in large-scale neural network optimization; conservative steps slow convergence, while… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Generative modeling of intrinsically disordered protein regions by reinforcing sparse autoencoder features | Intrinsically disordered protein regions (IDRs) play central roles in cellular processes such as transcriptional regulation, signal tr… | ↗Not briefed |
arxiv arxiv-ai | 1mo | DMAD: Distribution Matching as Adversarial Distillation for Fast Visual Generation | Distribution Matching Distillation (DMD) trains a few-step student from the difference between separately estimated target and student… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Higher-Order Molecular Grammars for Generative and Foundation Models in Chemistry | Molecular learning models are strongly shaped by their underlying representations. Yet standard sequential and graph formalisms strugg… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Decoding Looped Transformers Better for (Almost) Free | Looped Transformers achieve parameter efficiency by repeatedly executing a shared block across recurrent loops. Each loop yields an in… | ↗Not briefed |
hn hn-ai | 1mo | Show HN: Breadcrumb – A flight recorder and context manager for your AI (Mac)🔧 Ollama🔧 Gemma 3 | Hi HN, Justin here!<p>Breadcrumb records everything you do on your Mac (screen + meetings + AI transcripts + what you and your AI deci… | ↗Not briefed |
arxiv arxiv-ai | 1mo | SoftServe: A Scalable Quasi-Newton Method for Deep Learning | Quasi-Newton (QN) methods have long been among the most effective methods for large-scale unconstrained convex optimization. Two obsta… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Generative Cinematographer: Composing Camera and Object Motion in 3D | Current controllable video generation systems often rely on 2D motion trajectories or sparse drag signals for object motion. These con… | ↗Not briefed |
bluesky bluesky-ai | 1mo | it would be pretty cool to just have an agent go search for the cheapest airline tickets for you and present the options, but how can you e… | it would be pretty cool to just have an agent go search for the cheapest airline tickets for you and present the options, but how can… | ↗Not briefed |
arxiv arxiv-ai | 1mo | From Gradients to Capabilities: Understanding Multi-Teacher On-Policy Distillation | Multi-teacher on-policy distillation (MOPD) aims to combine the strengths of RL-trained teachers in a single student, but how teacher… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Effective Resistance and Graph Neural Network Reliability in Tissue-Specific Interactomes | Protein function annotation needs to know which predictions to distrust, not only what a model predicts. We ask whether tissue-specifi… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Have you been chiming in on AI discourse despite having no clue what linear algebra is? Let the AI version of me explain it to you. Claude… | Have you been chiming in on AI discourse despite having no clue what linear algebra is? Let the AI version of me explain it to you. C… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Every Ablation Is a Dose: Counterweights and the Semblance of Self-Repair | Ablate a component of a language model, and other components often appear to adjust and compensate. This phenomenon, termed self-repai… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Watch, Infer, Coordinate: Inferring Robot Partner Constraints for Zero-Shot Coordination | Robots operating in the physical world will increasingly need to coordinate with other robots, particularly in manipulation tasks wher… | ↗Not briefed |
hn hn-ai | 1mo | Worktrunk: A CLI for Git worktree management, designed for AI agent workflows | — | ↗Not briefed |
bluesky bluesky-ai | 1mo | these are not scientific conclusions | these are not scientific conclusions | ↗Not briefed |
bluesky bluesky-ai | 1mo | This is already happening with brands. Ever Google help on something, and find a 6 year old reddit thread with brand new comments recommend… | This is already happening with brands. Ever Google help on something, and find a 6 year old reddit thread with brand new comments reco… | ↗Not briefed |
arxiv arxiv-ai | 1mo | AutoCompact: Learning When to Compact Context in Long-Horizon Coding Agents | Coding agents solve repository-level software engineering tasks through long trajectories of code inspection, search, editing, and tes… | ↗Not briefed |
bluesky bluesky-ai | 1mo | And AI is going to make this problem SO MUCH WORSE. Another thing that AI is going to have to break is double-blind reviews. That's a grea… | And AI is going to make this problem SO MUCH WORSE. Another thing that AI is going to have to break is double-blind reviews. That's a… | ↗Not briefed |
bluesky bluesky-ai | 1mo | he's probably using dictation too which raises the question of how any LLM (not just gork) handles his garbage-stream way of talking. also… | he's probably using dictation too which raises the question of how any LLM (not just gork) handles his garbage-stream way of talking.… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Outrageous. OpenAI has reportedly fired three safety researchers for sharing information with an outside AI safety group. This looks like … | Outrageous. OpenAI has reportedly fired three safety researchers for sharing information with an outside AI safety group. This looks… | ↗Not briefed |
arxiv arxiv-ai | 1mo | DuoMind: Enabling Distributed Multi-Robot Coordination with Semantic Communication | Vision-language models (VLMs) and vision-language-action models (VLAs) have recently driven rapid progress in general-purpose robots,… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Today in nonconsensual shit being done to us by tech and media: Audible is ramming genAI LLM text to speech as a "feature" on audiobooks! … | Today in nonconsensual shit being done to us by tech and media: Audible is ramming genAI LLM text to speech as a "feature" on audiobo… | ↗Not briefed |
arxiv arxiv-ai | 1mo | When Do Intrinsic Rewards Lead to Exploration? | Intrinsic rewards are designed to guide exploration in reinforcement learning by assigning value to an agent's experience, for example… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Muon meets Tamed Langevin: Momentum Preconditioning beyond Convex and gradient-Lipschitz Potentials | We consider the problem of sampling from Gibbs distributions on matrix spaces whose potential energies are neither convex nor globally… | ↗Not briefed |
hn hn-ai | 1mo | In-House Counsel Is About to Get Good | — | ↗Not briefed |
rss techcrunch-ai | 1mo | Opus 5.5 loves to tell you ‘this matters’ (and other AI writing tells) | Opus 5.5’s biggest tell is the word “dependable,” which pops up 23 times more often than in human samples. | ↗Not briefed |
arxiv arxiv-ai | 1mo | From Knowledge Access to Source Learning: Developing Source-Specific Competence | Large language model (LLM) agents increasingly rely on persistent external sources to solve sequences of knowledge-intensive tasks. Ex… | ↗Not briefed |
bluesky bluesky-ai | 1mo | As if Stephen Miller’s Goebbels impression wasn’t enough, Trump now has the LLM that once called itself “MechaHitler” advising him. All the… | As if Stephen Miller’s Goebbels impression wasn’t enough, Trump now has the LLM that once called itself “MechaHitler” advising him. Al… | ↗Not briefed |
rss arstechnica-tech | 1mo | Memory executives expect RAM shortage to continue through 2028 | Prices for memory sold for 2027 "are much higher than 2026 prices." | ↗Not briefed |
arxiv arxiv-ai | 1mo | Faynt: Scaling and Optimizing Policies for Competitive Melee | We introduce Faynt, a family of 10M- and 75M-parameter Transformer policies for Super Smash Bros. Melee, each controlling all 26 chara… | ↗Not briefed |
hn hn-ai | 1mo | ExplainDB: A Database System Built for Understandability | — | ↗Not briefed |
bluesky bluesky-ai | 1mo | I think there's a fundamental misconception here; LLM fiction is not "robot, write me a book." It's working up an outline in tandem with th… | I think there's a fundamental misconception here; LLM fiction is not "robot, write me a book." It's working up an outline in tandem wi… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Keyword Harnesses Fail Open: A Cheap Diagnostic Ladder for Tool-Use Claims in Small Language Models | Keyword-matching benchmarks can credit small models for tool use they never perform. We document such a false positive in a matched-ar… | ↗Not briefed |
hn hn-ai | 1mo | Show HN: I built an app that scours the internet to create a daily news briefing | — | ↗Not briefed |
bluesky bluesky-ai | 1mo | Because another person can spend a couple hours or days fleshing out an idea and you can spend $1 to read it fresh? The issue is mostly tha… | Because another person can spend a couple hours or days fleshing out an idea and you can spend $1 to read it fresh? The issue is mostl… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Genuinely impressive digital history application of LLM: an unsolved - and untranscribed - enciphered Napoleonic era letter was broken usin… | Genuinely impressive digital history application of LLM: an unsolved - and untranscribed - enciphered Napoleonic era letter was broken… | ↗Not briefed |
bluesky bluesky-ai | 1mo | The promise of the LLM seems to be that every day your job will be a little worse until one day you can be laid off. | The promise of the LLM seems to be that every day your job will be a little worse until one day you can be laid off. | ↗Not briefed |
arxiv arxiv-ai | 1mo | Finetuning with Sampling: SFT Learns Better Than You Think | Introducing new capabilities to frontier models has long been the goal of posttraining, which predominantly employs supervised finetun… | ↗Not briefed |