AI News Today

Article URL: https://mrkt30.com/was-hugging-face-breached-by-ai-agents/ Comments URL: https://news.ycombinator.com/item?id=48989749 Points: 2 # Comments: 0

Meta has open-sourced Astryx, the React and StyleX design system it ran internally for eight years across 13,000+ apps. It ships 150+ accessible components, seven themes, dark mode, templates, and an agent-ready CLI under MIT — with React 19+ required. The post Meta Open-Sources Astryx: An Agent-Ready React Design System With 150+ Accessible Components, Seven Themes, and a CLI appeared first on MarkTechPost .

Get the best LG TV deal. Save 17% on the LG 100-Inch Class QNED evo AI QNED84B at Amazon.

Article URL: https://qwen.ai/blog?id=qwen-image-3.0 Comments URL: https://news.ycombinator.com/item?id=48989701 Points: 56 # Comments: 27

Article URL: https://medium.com/@klaudibregu/the-local-ai-lie-weve-all-been-sold-6d3f9373de5e Comments URL: https://news.ycombinator.com/item?id=48989678 Points: 2 # Comments: 0

Article URL: https://getwireai.com Comments URL: https://news.ycombinator.com/item?id=48989673 Points: 1 # Comments: 0

You all know the drill: you want to write prose using an AI because it can draft faster than you can type your ideas out. Next thing that happens, you find yourself in an editing loop: change this, fix there, explain your world in more detail. The chat window scroll bar shrinks to 2 pixels and the chapter gets worse every turn. You wonder if your prompts got bad or if the AI has forgotten your world. What actually happens is the context window starts working against you. A context window works f

Article URL: https://ai-song.net/ Comments URL: https://news.ycombinator.com/item?id=48989660 Points: 2 # Comments: 0

Article URL: https://aichatspot.online/ai-breakthroughs-guide-the-ultimate-overview-of-july-2026-innovations/ Comments URL: https://news.ycombinator.com/item?id=48989653 Points: 1 # Comments: 0

Find the best robot lawn mower deal. Save 22% on the roborock RockMow X120H at Amazon.

Rhoda AI is training robots with hundreds of millions of internet videos instead of lab data. Here's how its Direct Video Action model works.

Article URL: https://www.nextplatform.com/cloud/2026/07/20/microsoft-taps-amd-for-at-scale-ai-cpu-and-gpu-clusters/5275161 Comments URL: https://news.ycombinator.com/item?id=48989620 Points: 1 # Comments: 0

Article URL: https://github.com/alexzeitler/aceto Comments URL: https://news.ycombinator.com/item?id=48989614 Points: 1 # Comments: 0

<p> One identity layer for your customers, APIs, and agents </p> <p> <a href="https://www.producthunt.com/products/monocloud-for-startups-free-for-1-year?utm_campaign=producthunt-atom-posts-feed&utm_medium=rss-feed&utm_source=producthunt-atom-posts-feed">Discussion</a> | <a href="http://www.producthunt.com/r/p/1202223?app_id=339">Link</a> </p>

<h1> The Sad State of SWE Resumes (and How to Actually Land a Job in 2026) </h1> <p>We've all been there, waking up early, applying for jobs, trying to be among the first to apply, even when the role doesn't match. And now with powerful LLMs, you can tailor your resume for roles you aren't qualified for. Even if the gap is 100%, somehow you end up building a resume for it.</p> <p>Recruiters are overwhelmed by the volume of applications. Some say they can easily recognise AI-generated resumes, bu

<p>Across my previous articles, I’ve kept coming back to one central argument:</p> <p>Once UI rendering no longer owns the entire data flow, State, Derived State, Effects, and Async Work can no longer be carelessly compressed into the lifecycle of a single component.</p> <p>At first, this might sound like a frontend architecture problem:</p> <ul> <li>Components accumulating too much local state</li> <li>Effects becoming overloaded with unrelated side effects</li> <li>Async work calling <code>set

<div class="ltag__link--embedded"> <div class="crayons-story "> <a href="https://dev.to/owasp/owasp-cornucopia-v34-eop-help-pages-and-phantom-b-3lal" class="crayons-story__hidden-navigation-link">OWASP Cornucopia v3.4: EoP help pages and PHANTOM-B</a> <div class="crayons-story__body crayons-story__body-full_post"> <div class="crayons-story__top"> <div class="crayons-story__meta"> <div class="crayons-story__author-pic"> <a class="crayons-logo crayons-logo--l" href="/owasp"> <img alt="OWASP® Found

<pre style='white-space:pre-wrap;width:81ex'>fix: make Copilot token exchange 403 errors actionable (#112025) * fix: explain GitHub Copilot 403 recovery * fix: keep control UI log hint internal * fix: clarify log viewing hint</pre>

Article URL: https://docs.vulkan.org/tutorial/latest/AI_Assisted_Vulkan/introduction.html Comments URL: https://news.ycombinator.com/item?id=48989489 Points: 2 # Comments: 0

<h1> My Local AI Assistant Got Worse When I Remembered Too Much </h1> <p>I moved a personal AI assistant onto a small local model last week and immediately hit a boring problem: the model was fine, but my memory layer was not.</p> <p>The old version persisted raw conversation history and replayed it back into the prompt. That worked well enough with hosted models. Then I pointed the same app at a local OpenAI-compatible server running Qwen3-4B-4bit through Swama on the Mac mini.</p> <p>After 196

<p>Large language models are very good at producing fluent text. That does not mean every fluent rewrite is a faithful rewrite.</p> <p>While building <a href="https://tryparaphraser.com/" rel="noopener noreferrer">Paraphraser AI</a>, I kept coming back to a deceptively simple requirement:</p> <blockquote> <p>Change the wording without changing the claim.</p> </blockquote> <p>That requirement becomes difficult as soon as users ask for stronger edits, a different tone, several output versions, or

I'm an indie developer and use both macOS and Windows for development. Most of my indie game work is done in Godot on Windows. I started using AI coding agents a few months ago, and one problem kept bothering me: Windows doesn't have a simple and reliable equivalent to tmux or screen. If I accidentally closed a terminal window, the agent session running inside it was gone too. I tried tmux through MSYS2. It sort of worked, but it would sometimes stop responding for no obvious reason. That's why

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

Enterprises evaluating Chinese open-weight models this month face a question that has nothing to do with benchmarks: whether using one will still be straightforward in a year. Moonshot AI’s Kimi K3 arrived on July 16 as the largest open-weight model yet released, and within days it had reopened a policy argument in Washington that had […] The post Chinese open-weight models are cheap. Washington is deciding what that costs. appeared first on AI News .

<pre style='white-space:pre-wrap;width:81ex'>fix(media): bound stalled response headers by default (#103209) * fix(agent-tools): bound remote media fetches with a 15-min total timeout Apply the shared bounded-media ceiling (15 min, matching the canonical precedent in PR #103266) to the three agent-tool remote-media callers (image, image-generate, pdf) via the new REMOTE_MEDIA_TIMEOUT_MS constant, forwarded through loadWebMedia / loadWebMediaRaw's timeoutMs seam. Bounds endpoints that never respo

And why? I'd genuinely like to understand who is on which side and why. I'd also like to know, if you use one of them, what kind of subscription plan you have. (For example, I use Claude (max) [...]) Comments URL: https://news.ycombinator.com/item?id=48989357 Points: 4 # Comments: 4

<!-- SC_OFF --><div class="md"><p>Hi everyone,</p> <p>I'm working on extracting the hierarchical structure of long PDF documents (legal/regulatory text, lots of numbered sections) and would like to gather some feedback on my approach before committing to it.</p> <p><strong>What I've done so far:</strong> I render each PDF page to an image and run it through <a href="https://huggingface.co/baidu/Unlimited-OCR">Baidu's DeepSeek-OCR model</a>. It returns each detected block with a bounding box <cod

<pre style='white-space:pre-wrap;width:81ex'>refactor(media): introduce neutral MediaFact owner with identity-preserving adapters (#112197) * refactor(media): introduce neutral MediaFact owner with identity-preserving adapters Media-facts program PR 1 (audit-frozen plan): canonical MediaFact ownership and normalization/projection helpers in a neutral core media module; legacy channel/plugin/agent payload builders become adapters that project exactly their declared input fields (structural richer

NVIDIA has released Cosmos 3 Edge, a 4-billion-parameter open world model built to run on-device. It helps robots and vision AI agents understand surroundings, reason in real time, and generate robot actions locally. The Cosmos 3 family included Cosmos 3 Nano (16B) and Cosmos 3 Super (64B) shipped on May 31, 2026 at GTC Taipei. […] The post NVIDIA Releases Cosmos 3 Edge: A 4B-Parameter Open World Model That Reasons and Generates Robot Actions On-Device appeared first on MarkTechPost .

Article URL: https://www.tomshardware.com/tech-industry/artificial-intelligence/trump-administration-reportedly-reviving-push-to-ban-chinese-ai-models-following-kimi-k3-launch-citing-cybersecurity-concerns-downloadable-open-weights-could-make-an-outright-u-s-ban-nearly-impossible-to-enforce-amid-growing-adoption Comments URL: https://news.ycombinator.com/item?id=48989230 Points: 3 # Comments: 1

Researchers at Sysdig have linked a second attack on the same Langflow server to JADEPUFFER, the AI-agent-driven operator it first documented earlier this month. The same operator has now been spotted deploying ENCFORGE, a new compiled Go ransomware designed to encrypt model weights, vector indexes, training datasets, and other AI infrastructure files across the host filesystem. The entry

<pre style='white-space:pre-wrap;width:81ex'>fix(agents): drop orphaned thinking blocks when NO_REPLY text leaves an assistant message with only thinking content (#99620) (#99772) * fix(agents): drop silent reasoning-only replay turns Prevent NO_REPLY cleanup from leaving signed reasoning that merges into the next assistant tool-use turn and bricks strict-provider sessions. Related: #99620 Co-authored-by: SunnyShu <shu.zongyu@xydigit.com> * docs: refresh transcript hygiene map Regenerate the doc

Article URL: https://tryparaphraser.com/ Comments URL: https://news.ycombinator.com/item?id=48989131 Points: 2 # Comments: 0

<!-- SC_OFF --><div class="md"><p><strong>TL;DR:</strong> I’m reproducing the trait-persistence result from <a href="https://arxiv.org/abs/2606.24014">arXiv:2606.24014</a> on one RTX 3090. Before I can test persistence I need to <em>install</em> a trait via RL — and my GRPO run moves the trait only <strong>+2.4 points</strong> (95% CI [+0.2, +4.8]) when I need ~+15. Training is mechanically healthy and I’ve ruled out the obvious culprits. Looking for advice from people who’ve done small-scale RL

Article URL: https://brightsec.com/research/detecting-ansi-escape-sequence-injection-in-mcp-servers-with-dast/ Comments URL: https://news.ycombinator.com/item?id=48989006 Points: 1 # Comments: 0

<p> All-in-one AI Platform for Teams </p> <p> <a href="https://www.producthunt.com/products/aymo-ai?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/1202119?app_id=339">Link</a> </p>

Article URL: https://github.com/vijaym2k6/SteerPlane Comments URL: https://news.ycombinator.com/item?id=48988992 Points: 1 # Comments: 0

Presented by Atlassian Most companies are approaching AI adoption backwards by optimizing how individuals use AI instead of how teams work together, said Dr. Molly Sands, head of the Teamwork Lab at Atlassian, during a fireside chat with VentureBeat senior technology contributor Sam Witteveen at VB Transform 2026 . Sands leads a team of behavioral scientists and psychologists who study how AI is reshaping the way people work together, using those findings to help organizations redesign how work

Presented by Atlassian Most companies are approaching AI adoption backwards by optimizing how individuals use AI instead of how teams work together, said Dr. Molly Sands, head of the Teamwork Lab at Atlassian, during a fireside chat with VentureBeat senior technology contributor Sam Witteveen at VB Transform 2026 . Sands leads a team of behavioral scientists and psychologists who study how AI is reshaping the way people work together, using those findings to help organizations redesign how work

OpenAI and Hugging Face share early findings from a security incident during AI model evaluation, highlighting advanced cyber capabilities and lessons for defenders.