AI News Today

arXiv:2604.09523v3 Announce Type: replace Abstract: Training reinforcement-learning agents for cyber defense requires an environment that reflects the operational setting: noisy, partial observations, several defenders coordinating across a network, and an adaptive adversary realized through self-play. We present NetForge RL, a multi-agent environment for this setting on procedurally generated enterprise and operational-technology (OT) networks. A red agent compromises hosts with partial observa

arXiv:2605.12765v3 Announce Type: replace Abstract: Large Language Models memorize vast amounts of training data, raising concerns regarding privacy, copyright infringement, and safety. Machine unlearning seeks to remove the influence of a targeted forget set while preserving model performance, ideally approximating a model retrained from scratch without the forget set. Existing approaches aim to achieve this by updating model parameters via gradient-based methods. However, these updates are com

arXiv:2605.21107v2 Announce Type: replace Abstract: We study constrained online convex optimization with adversarial time-varying constraints. At each round the learner acts before observing the loss and constraint, and is compared with the best fixed action satisfying all constraints in hindsight. The goal is to obtain minimax-optimal regret while controlling cumulative constraint violation (CCV). Prior algorithms achieved $O(\log T)$ regret with $O(\sqrt{T\log T})$ CCV for strongly convex loss

arXiv:2605.22432v2 Announce Type: replace Abstract: Modern deep learning commonly relies on AdamW with prescribed learning rate schedules, but recent works challenge both components: Schedule-Free optimization removes explicit schedules via iterate averaging, and Muon improves the update geometry by orthogonalizing momentum for matrix parameters. Despite Muon's strong empirical performance, its underlying mechanism remains partially understood. We study Muon through the river-valley loss landsca

arXiv:2605.28198v2 Announce Type: replace Abstract: Existing approaches for synthetic tabular data generation are based on either purely generative models or LLMs, both of which struggle with data heterogeneity, logical consistency, rare-event coverage, and robustness in low-data regimes. In this paper, we propose a hierarchical hybrid top-down and bottom-up (H-TDBU) framework that decouples semantic structures from stochastic texture. In the top-down path, structure-driven logical constraints a

Article URL: https://millfolio.app/blog/millwright-ui-layers/ Comments URL: https://news.ycombinator.com/item?id=48916095 Points: 2 # Comments: 0

PM frames sweeping new regulations as the equivalent of labour movement touchstones like winning a minimum wage

Article URL: https://video-editor.ai-creator.top/ Comments URL: https://news.ycombinator.com/item?id=48915771 Points: 1 # Comments: 0

Article URL: https://www.minister.industry.gov.au/t-ayres/media/ai-australias-interests Comments URL: https://news.ycombinator.com/item?id=48915705 Points: 1 # Comments: 0

Article URL: https://www.kadoa.com/blog/ai-agents-hype-vs-reality Comments URL: https://news.ycombinator.com/item?id=48915700 Points: 1 # Comments: 0

<p>Coherent decision-making and internal accountability are critical to meeting this manic moment</p><ul><li><p><a href="https://www.theguardian.com/technology/2026/jul/14/anthony-albanese-promises-fast-track-approvals-for-datacentres-to-shore-up-ai-investment">Anthony Albanese promises fast-track approvals for datacentres to shore up AI investment</a></p></li></ul><p>The University of Sydney was the natural setting for Anthony Albanese to lay out his vision for how Australia should confront the

Article URL: https://celestial-llm-mirror.vercel.app/ Comments URL: https://news.ycombinator.com/item?id=48915548 Points: 1 # Comments: 0

Article URL: https://www.cnn.com/2026/07/14/tech/ai-cybersecurity-clearing-house-white-house Comments URL: https://news.ycombinator.com/item?id=48915488 Points: 1 # Comments: 0

Article URL: https://pullcard.sakimyto.com Comments URL: https://news.ycombinator.com/item?id=48915464 Points: 1 # Comments: 0

Article URL: https://perseus.observer/blog/built-perseus-vault-obsidian-wasnt-cutting-it/ Comments URL: https://news.ycombinator.com/item?id=48915435 Points: 1 # Comments: 0

Article URL: https://gomoku.w3cub.com/ Comments URL: https://news.ycombinator.com/item?id=48915369 Points: 1 # Comments: 0

<p> Email for AI agents. They sign up, send and reply. </p> <p> <a href="https://www.producthunt.com/products/nitrosend?utm_campaign=producthunt-atom-posts-feed&utm_medium=rss-feed&utm_source=producthunt-atom-posts-feed">Discussion</a> | <a href="https://www.producthunt.com/r/p/1196771?app_id=339">Link</a> </p>

The good kind of snack. 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/grigoriitropin/vehir-platform Comments URL: https://news.ycombinator.com/item?id=48914954 Points: 2 # Comments: 0

Article URL: https://bunnyandcloud.com/ Comments URL: https://news.ycombinator.com/item?id=48914764 Points: 2 # Comments: 0

Lightspeed is in discussions to lead the funding round, sources say.

Article URL: https://www.privacyguides.org/news/2026/07/13/the-next-version-of-metas-ai-glasses-will-activate-the-camera-without-the-camera-indicator-light/ Comments URL: https://news.ycombinator.com/item?id=48914711 Points: 3 # Comments: 0

<p><em>How a distributed, event-sourced issue tracker built with developer ergonomics in mind may have a role to play in the next generation of agentic workflows</em></p> <h2> Vision drift </h2> <p>Harness engineering has recently popularized the idea of containing <strong><em>architectural drift</em></strong> in agentic workflows. What might be missing in the discussion is a similar issue on a higher level - <strong><em>vision drift</em></strong>.</p> <p>By vision drift I mean that the implemen

To serve the 397B-parameter Qwen 3.5 Mixture-of-Experts (MoE) model on Ironwood TPUs, engineers developed a modular JAX/Pallas optimization stack that achieved up to a 4.7x inference speedup for prefill-heavy workloads. The team bypassed severe hardware sharding constraints by deploying a hybrid Data Parallelism and Expert Parallelism (DP+EP) topology, paired with custom low-level communication fusions like a hierarchical reduce-scatter to optimize cross-device token routing. Finally, by executi

I run a boutique investment bank, and every year I would receive 1 - 3 internship requests, but this year I received 50. I started to realize that Ai was replacing interns faster than any other worker because it was mostly administrative work. Internships really gave me invaluable experience, so I wanted to create a de-centralized internship with the first year content focused on a de-centralized Ai protocol called Bittensor (TAO). There are 8 weekly assignments which are scored and reviewed by

Humanity briefly regains critical thinking skills Comments URL: https://news.ycombinator.com/item?id=48914572 Points: 3 # Comments: 2

<p>This is the final installment of the "Claude Code environment" series that ran all the way through <a href="https://zenn.dev/bokuwalily/articles/hook-latency-profiling" rel="noopener noreferrer">profiling hook latency</a>.</p> <p>Up to now I've built the parts one at a time — "memory," "skills," "cost," "self-healing." This finale does two things. First, it lays those parts back out on a single map and puts into words <strong>the connections between the parts that never made it into the indiv

Today we're publishing the AI Use-Case Library: 159 real, named AI deployments across 21 industries, with the tools, vendors, and reported outcomes on 77 of them. Six were halted or reversed, and those might be the most useful entries in the file. It's free, no signup, and built to search before you spend a budget or write a pitch. This edition is the tour: what's working, what got pulled back, and why a precedent file matters right now.

Visual generative models (e.g., diffusion models) typically operate in compressed latent spaces to balance training efficiency and sample quality. In parallel, there has been growing interest in leveraging high-quality pre-trained visual representations—either by aligning them inside VAEs or directly within the generative model. However, adapting such representations remains challenging due to fundamental mismatches between understanding-oriented features and generation-friendly latent spaces. R

Inkling from Thinking Machines is now available on AI Gateway. Inkling is a broad generalist model, trained across agentic, reasoning, coding, instruction-following, factuality, vision, and audio tasks rather than optimized for a single domain. The model also supports controllable thinking effort. To use Inkling, set model to thinkingmachines/inkling in the AI SDK : AI Gateway provides a unified API for calling models, tracking usage and cost, and configuring retries, failover, and performance o

Retrieval-augmented generation (RAG) enhances large language models (LLMs) with external knowledge but still suffers from long contexts and disjoint retrieval–generation optimization. In this work, we propose CLaRa (Continuous Latent Reasoning), a unified framework that performs embedding-based compression and joint optimization in a shared continuous space. To obtain semantically rich and retrievable compressed vectors, thereby reducing the document length fed into the generator, we introduce S

Large Language Models (LLMs) are increasingly deployed to autonomously solve real-world tasks. A key ingredient for this is the LLM Function-Calling paradigm, a widely used approach for equipping LLMs with tool-use capabilities. However, an LLM calling functions incorrectly can have severe implications, especially when their effects are irreversible, e.g., transferring money or deleting data. Hence, it is of paramount importance to consider the LLM’s confidence that a function call solves the ta

GitHub Tools now has first-class support for Vercel Connect through the new @github-tools/sdk/connect subpath. Instead of storing a long-lived personal access token, your agent mints short-lived, scoped GitHub tokens at runtime from a connector. There is no secret to store, rotate, or leak. Attach a GitHub connector to your project and pick a preset: Scopes from presets: Presets such as code-review , issue-triage , and maintainer map to Connect scopes automatically, so tokens carry only the perm

<p>If your database documentation is always behind production, this is for you.</p> <p><strong>SchemaCrawler Scribe</strong> generates structured database documentation directly from live schema metadata, using <strong><a href="https://github.com/GoogleCloudPlatform/knowledge-catalog/blob/d44368c15e38e7c92481c5992e4f9b5b421a801d/okf/SPEC.md" rel="noopener noreferrer">Google Open Knowledge Format (OKF)</a></strong>.</p> <p>The result is documentation that works for both developers and AI agents,

The Verge's David Pierce kicks off our new series on staying productive in the AI era — starting with why you should stop trying to stay ahead

Article URL: https://lexifina.com/blog/legal-ai-not-a-coding-agent-with-scaffolding Comments URL: https://news.ycombinator.com/item?id=48914472 Points: 4 # Comments: 0

Comments URL: https://news.ycombinator.com/item?id=48914471 Points: 8 # Comments: 4

Article URL: https://jordangreen.bearblog.dev/why-organizing-knowledge-in-the-age-of-ai-sucks-and-what-i-built/ Comments URL: https://news.ycombinator.com/item?id=48914439 Points: 3 # Comments: 1

I wanted an ASCII animation for one of my projects but couldn't find any collection or library for anything like that, so I cooked this up over a weekend with some ideas, inspiration and Claude. Feel free to add your own patterns! Comments URL: https://news.ycombinator.com/item?id=48914428 Points: 1 # Comments: 0