AI News Today

arXiv:2603.21210v3 Announce Type: replace Abstract: Designing urban spaces that provide pedestrian wind comfort and safety requires time-resolved Computational Fluid Dynamics (CFD) simulations, but their current computational cost makes extensive design exploration impractical. We introduce WinDiNet (Wind Diffusion Network), a pretrained video diffusion model that is repurposed as a fast, differentiable surrogate for this task. Starting from LTX-Video, a 2B-parameter latent video transformer, we

arXiv:2603.23461v2 Announce Type: replace Abstract: We study reinforcement learning (RL) with linear function approximation in Markov Decision Processes (MDPs) satisfying \emph{linear Bellman completeness} -- a fundamental setting where the Bellman backup of any linear value function remains linear. While statistically tractable, prior computationally efficient algorithms are either limited to small action spaces or require strong oracle assumptions over the feature space. We provide a computati

arXiv:2603.29135v2 Announce Type: replace Abstract: Autonomous experimental systems are increasingly used in materials research to accelerate scientific discovery, but their performance is often limited by low-quality, noisy data. This issue is especially problematic in data-intensive structure-property learning tasks such as Image-to-Spectrum (Im2Spec) and Spectrum-to-Image (Spec2Im) translations, where standard active learning strategies can mistakenly prioritize poor-quality measurements. We

arXiv:2604.00860v4 Announce Type: replace Abstract: Reinforcement learning has become a central post-training paradigm for improving LLM and agent capabilities. Yet existing RL post-training methods share a common blind spot: they construct local learning signals from sampled trajectories, rewards, or feedback-conditioned targets, then update the policy without explicitly verifying whether the resulting policy outperforms its predecessor. Optimizing these local signals does not necessarily produ

arXiv:2604.08492v2 Announce Type: replace Abstract: Previous work has shown that node embedding methods can produce different representations and downstream predictions across repeated training runs, even when trained on the same data with identical hyperparameters. However, the role of embedding dimensionality in this instability remains poorly understood. In this work, we systematically analyze how embedding dimensionality affects the stability of embeddings from five widely used node embeddin

arXiv:2605.00553v3 Announce Type: replace Abstract: Large Language Model (LLM) Red-Teaming, which proactively identifies vulnerabilities of LLMs, is an essential process for ensuring safety. Finding effective and diverse attacks in red-teaming is important, but achieving both is challenging. Generative Flow Networks (GFNs) that perform distribution matching are promising methods, but they are notorious for training instability and mode collapse. In particular, unstable rewards in red-teaming acc

arXiv:2605.04853v3 Announce Type: replace Abstract: We propose HIN-LRI, a hybrid framework that augments a classical numerical solver with a neural operator trained to correct the solver's structured truncation error. A base low-regularity integrator provides a consistent first-order approximation to nonlinear dispersive PDEs, while a lightweight neural network, operating on a low-dimensional latent manifold, learns the residual defect that analytical methods cannot close. An explicit time-step

arXiv:2605.11554v2 Announce Type: replace Abstract: Task-agnostic structure proxies are often used to interpret why one pretraining corpus transfers better than another, but such explanations require the proxy to track the structure that matters for the downstream task. We test this requirement in a fixed pretraining-and-probing setup motivated by computationally bounded notions of learned structure, including epiplexity. The core question is whether a proxy ranking of two pretraining datasets m

arXiv:2605.24661v3 Announce Type: replace-cross Abstract: Despite remarkable progress on reasoning benchmarks, current LLM evaluation practice remains anchored to final-answer correctness, providing limited insight into how models reason, how reliably they behave under contextual variation, or how efficiently they reach conclusions. This paper proposes a unified multi-dimensional framework for measuring LLM reasoning quality from a behavioral perspective, operationalizing six theoretically groun

arXiv:2606.09052v3 Announce Type: replace-cross Abstract: Self-evolution offers a scalable path to stronger reasoning: a pretrained language model improves itself with only minimal external supervision. Yet existing methods either depend on extensively curated or teacher-generated training data, or, when the generator runs unsupervised, reward it by a difficulty heuristic that need not improve the solver. We introduce INFUSER, an iterative co-training framework with two co-evolving roles: a Gene

arXiv:2606.12634v2 Announce Type: replace-cross Abstract: Long-horizon tool-use reinforcement learning learns from outcome verification, but trajectory-level advantages are broadcast over reasoning, API, and answer tokens. Direct self-distillation can supply a denser signal, but in our experiments it can also destroy tool use by rehearsing teacher behavior without identifying which actions the verifier rewards. We introduce Sibling-Guided Credit Distillation (SGCD), which uses distillation for b

arXiv:2606.13239v2 Announce Type: replace-cross Abstract: Existing computer-use agents remain fundamentally limited in professional software manipulation: GUI-based agents suffer from fragile visual grounding and long-horizon error accumulation, while API-basedapproaches struggle with heterogeneous protocols and inaccessible commercial interfaces. In this work,we identify the Component Object Model (COM) as a unified executable abstraction, proposing COM-as-Action: a new paradigm that reframes p

<pre style='white-space:pre-wrap;width:81ex'>fix(security): warn on agent skill MCP boundary drift (#98352) Summary: - Merged fix(security): warn on agent skill MCP boundary drift after ClawSweeper review. Automerge notes: - No ClawSweeper repair was needed after automerge opt-in. Validation: - ClawSweeper review passed for head ab3c29ef4c731dfdbff75f2f5c7ec65c63ec195d. - Required merge gates passed before the squash merge. Prepared head SHA: ab3c29ef4c731dfdbff75f2f5c7ec65c63ec195d Review: http

Just came across a great Medium article that shows how to automate the entire process of discovering newly added CISA KEV vulnerabilities and generating Sigma detection rules using AI. Instead of manually tracking new CVEs and writing detections from scratch, the workflow automatically: Identifies newly published KEV vulnerabilities Generates Sigma detection rules with AI Maps them to MITRE ATT&CK Distributes the results to Google Sheets, Slack, email, and your SIEM If you're a SOC analyst, dete

Article URL: https://www.euronews.com/my-europe/2026/06/30/america-can-switch-off-the-worlds-ai-europe-must-switch-gears-before-its-too-late Comments URL: https://news.ycombinator.com/item?id=48741943 Points: 22 # Comments: 7

The Gemini macOS app now has access to Google's Spark agentic AI assistant.

Article URL: https://news.nate.com/view/20260701n15527?mid=n1101 Comments URL: https://news.ycombinator.com/item?id=48741900 Points: 2 # Comments: 0

<pre style='white-space:pre-wrap;width:81ex'>fix(heartbeat): scope commitment fan-out prompts (#98169) * fix(heartbeat): scope commitment fan-out prompts * fix(heartbeat): isolate commitment fan-out runs * fix(heartbeat): isolate commitment fan-out runs --------- Co-authored-by: Benjamin Badejo <ben@benbadejo.com> Co-authored-by: Peter Steinberger <steipete@gmail.com></pre>

It’s our goal for AWS to be the most secure place to run any workload, and in support of that we’ve been deeply investing in security across our services since AWS's inception more than two decades ago. Our AI services like Amazon Bedrock are built on this foundation and with the same focus.

Article URL: https://buttondown.com/maiht3k/archive/how-to-talk-about-ai-without-adding-to-the/ Comments URL: https://news.ycombinator.com/item?id=48741809 Points: 3 # Comments: 0

Article URL: https://www.empirical.health/blog/llm-scaling-laws-hold-for-sensor-data/ Comments URL: https://news.ycombinator.com/item?id=48741774 Points: 3 # Comments: 0

Anthropic said it would begin restoring access to the Fable on July 1.

Article URL: https://www.gizmoweek.com/xiaomi-clip-on-earbuds/ Comments URL: https://news.ycombinator.com/item?id=48741603 Points: 3 # Comments: 0

Article URL: https://smartcontractauditor.ai/ Comments URL: https://news.ycombinator.com/item?id=48741579 Points: 2 # Comments: 0

Wayve’s offering is part of a growing trend of AI startups using employee tenders as a strategic tool to attract and retain talent.

Antarctica's Blood Falls are even weirder than they look. ScienceAlert stories are written, fact-checked, and edited by humans, never generated by AI. Don't miss a story, subscribe here.

I realized last year that I needed a better shared substrate for all the parallel coding (and other) agents I run if I wanted all my ADHD agentically-engineered dreams to come true. It started off as a simple Kanban board for agents with a crawler and rag, but it's long since evolved into a high performance, scalable, multi-user, self-hostable agent memory system built on SurrealDB that I can't live without. It's the thing that's helped me go absurdly parallel without just turning into a slop ca

Article URL: https://duckduckgo.com/?q=site%3Abandcamp.com+covers.pics Comments URL: https://news.ycombinator.com/item?id=48741550 Points: 3 # Comments: 1

Anthropic will start its users' access to Mythos and Fable tomorrow, July 1.

<pre style='white-space:pre-wrap;width:81ex'>fix(agents): estimate harness role sizes in context guard char estimator (#97928) The in-loop context-overflow guard (installToolResultContextGuard) sizes transcript messages via estimateMessageChars, which only handled user, assistant, and toolResult roles. Harness roles (bashExecution, compactionSummary, branchSummary, custom) fell through to a flat 256-char return, causing the guard to undercount summary- and bash-dominated context by orders of mag

<pre style='white-space:pre-wrap;width:81ex'>fix(auto-reply): stop level directives from eating the next message word (#97929) matchLevelDirective consumed the token after a level directive (/think, /verbose, /trace, /fast, /reasoning, /elevated) as the level argument unconditionally, and extractLevelDirective stripped it from the body whether or not it was a valid level. So a message like "/verbose explain quantum computing" reached the agent as "quantum computing", silently dropping the user's

<p>The AI company was forced earlier this month to suspend access to its Fable 5 and Mythos 5 models for all foreign nationals</p><p>Anthropic has said the US commerce department has lifted export controls on its Fable and Mythos AI models, less than ⁠three weeks after ⁠the company ​was ordered to suspend access to its most advanced AI models over national security risks.</p><p>“We’ll begin restoring access tomorrow,” Anthropic said in a statement on X late on Tuesday.</p> <a href="https://www.t

Article URL: https://www.microsoft.com/en-us/corporate-responsibility/topics/ai-economy-institute/reports/us-ai-adoption-2026-q1/ Comments URL: https://news.ycombinator.com/item?id=48741390 Points: 2 # Comments: 1

Huh?? ScienceAlert stories are written, fact-checked, and edited by humans, never generated by AI. Don't miss a story, subscribe here.

Article URL: https://github.com/arman-jalili/rigorix-oss Comments URL: https://news.ycombinator.com/item?id=48741332 Points: 3 # Comments: 0

If you had a choice between Fable 5 by Anthropic or Replit which would you use. Which one would make a better vibe coded applicaiton? Comments URL: https://news.ycombinator.com/item?id=48741331 Points: 3 # Comments: 0

Article URL: https://missionlocal.org/2026/06/s-f-supervisors-balk-at-362-page-city-code-rewrite/ Comments URL: https://news.ycombinator.com/item?id=48741316 Points: 2 # Comments: 0

Meet Agentic OS - your proactive AI assistant that seamlessly automates tasks, scheduling, and files Comments URL: https://news.ycombinator.com/item?id=48741277 Points: 1 # Comments: 0

<!-- SC_OFF --><div class="md"><p>Today is the day you (🫵!) get access to 80TB plus of data from over 30 astronomical surveys in one place. 4GB of RAM is enough even at Gaia Scale. Check out our writeup here:</p> <p><a href="https://huggingface.co/blog/hugging-science/multimodal-universe-hats">https://huggingface.co/blog/hugging-science/multimodal-universe-hats</a></p> <p>And a tutorial here <a href="https://asciinema.org/a/1259218">https://asciinema.org/a/1259218</a></p> </div><!-- SC_ON --> s

Article URL: https://s2.dev/blog/local-ai Comments URL: https://news.ycombinator.com/item?id=48741231 Points: 1 # Comments: 0