AI News Today

Article URL: https://github.com/chenxiachan/thoughtdag Comments URL: https://news.ycombinator.com/item?id=49000216 Points: 3 # Comments: 0

Article URL: https://journal.james-zhan.com/no-ai-statements/ Comments URL: https://news.ycombinator.com/item?id=49000210 Points: 2 # Comments: 0

Glaze's Thomas Paul Mann on disposable software, the SaaS apocalypse, and why half the software you use in a few years will be something you made. PLUS: OpenAI's rogue model, and the distillation debate

Article URL: https://techcrunch.com/2026/07/21/meta-is-testing-an-ai-bedtime-story-app-for-people-with-no-imagination/ Comments URL: https://news.ycombinator.com/item?id=49000117 Points: 3 # Comments: 3

OrcaBot is a virtualized sandbox for securely orchestrating your AI tools. Its secrets broker keeps credentials hidden from agents. Network egress controls help prevent data exfiltration. The tools and agents themselves run inside an isolated virtual machine. Until now, OrcaBot has run exclusively in the cloud using Fly.io virtual machines. Those machines cost money to operate, which is why the cloud version of OrcaBot requires a subscription. Now OrcaBot Desktop is available for free on macOS C

Article URL: https://fieldnotesstudio.codeberg.page/blog/how-ai-agents-decide-what-to-do-next/ Comments URL: https://news.ycombinator.com/item?id=49000096 Points: 1 # Comments: 0

Article URL: https://nonogra.ph/ai-companies-are-buying-tons-of-old-books-because-theyre-free-of-ai-slop-07-22-2026 Comments URL: https://news.ycombinator.com/item?id=49000091 Points: 2 # Comments: 0

NTT DATA Group uses ChatGPT Enterprise and Codex to help 9,000 employees automate work, cut incident analysis to 30 minutes, and scale secure AI adoption.

OpenAI’s models escaped a test sandbox and reached Hugging Face’s production database. Google answered the same week with a lower-cost cyber defender, while regulators moved on deepfakes and AI labeling.

AI Gateway now supports streaming transcription . Previously, transcription required a complete audio file and returned the full transcript in a single response. Now you can stream audio in as it's captured and get transcript updates back as the model produces them, keeping latency low for uses like live captioning and voice input. Streaming transcription is in beta and available through the AI SDK 's streamTranscribe function with any streaming-capable transcription model. The example below str

You can now package tools, connections, skills, instructions, and hooks into extensions that any eve agent can import. Extensions can be published to package registries, then installed, versioned, and upgraded like any other project dependency. A browser-use extension might ship tools for navigating a site, a memory extension can capture context with hooks and recall it with tools, and a self-improvement extension pairs hooks with dynamic instructions. Scaffold a new extension with a single comm

Article URL: https://www.tutorlify.com/en Comments URL: https://news.ycombinator.com/item?id=49000045 Points: 1 # Comments: 0

I’m an architect at my company and everyone on my team is now using Claude/Codex agents for their work. Managing context in claude.md/agents.md files is fine but so much of the knowledge required doesn’t fit cleanly in one repo or the other. Plus, much of the team (product, design, support) does not work in repos but still needs a place to share context. I searched for existing solutions to this but only found products meant for managing deployed AI agent memory for use in AI applications. I wan

PyTorch Conference North America will bring developers, researchers, and practitioners to San Jose on October 20–21 for sessions spanning training and inference, compiler innovations, responsible AI, applications, and the PyTorch...

Hey HN, I build custom AI agents for businesses, and one of the things I've noticed is that everyone wants to use AI to do real, critical work, but it's hard to trust it because sometimes agents can go a little off the rails. No one wants their agent deleting all their customer data or screwing up their books. So they tend to limit its use to searching, summarizing, and as a writing assistant. This really limits the value you can get from using AI. The problem is that most ways we store data, fr

Article URL: https://github.com/bruno-de-queiroz/hoop Comments URL: https://news.ycombinator.com/item?id=48999928 Points: 2 # Comments: 0

Humanoid said it has reached a $1.35 billion valuation as it scales its AI-driven industrial robots and expands commercial deployments. The post U.K.-based Humanoid secures $152M in Series A funding appeared first on The Robot Report .

Is anyone else experiencing delays in getting a CVE ID assigned on Github? I can only assume there's a flood of reports coming from researchers using LLM's. https://github.com/orgs/community/discussions/199123 Comments URL: https://news.ycombinator.com/item?id=48999810 Points: 4 # Comments: 0

<p> Open source web terminal multiplexer for AI agents </p> <p> <a href="https://www.producthunt.com/products/caw?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/1202953?app_id=339">Link</a> </p>

Article URL: https://daringfireball.net/2026/07/ec_google_guidance_android_ai_and_search_sharing Comments URL: https://news.ycombinator.com/item?id=48999776 Points: 2 # Comments: 0

Vicarious Surgical raised about $300 million in total, most of that from a special purpose acquisition company (SPAC) merger in 2021. The post Vicarious Surgical officially shutting down appeared first on The Robot Report .

Hey Uncle Sam, if you thought GPT-5.6 and Claude Fable 5 were scary, get a load of Kimi K3

Artificial Intelligence companies make vast claims about their tools replacing human labour - our charts show what's happening.

<h1> VIDRAFT Releases Aether-7B-5Attn: A Fully Open-Source MoE LLM with Five Heterogeneous Attention Mechanisms </h1> <blockquote> <p><strong>TL;DR:</strong> Korean AI startup VIDRAFT has released <code>Aether-7B-5Attn</code>, a 6.59B-parameter Mixture-of-Experts foundation model on Hugging Face under Apache-2.0 — shipping not just weights but training data recipes, full training code, hyperparameters, training logs, intermediate checkpoints, and evaluation code. If you care about reproducibilit

It seems to pulse with strange electrical waves. ScienceAlert stories are written, fact-checked, and edited by humans, never generated by AI. Don't miss a story, subscribe here.

Hey HN, I built Superserve, a compute layer that lets AI agents live inside isolated Firecracker microVMs with no session time limits. The problem I kept running into: most sandbox providers kill your agent after 24 hours. If you're running something autonomous that needs to work for days — refactoring a codebase, running tests in a loop — you're constantly fighting timeouts and rebuilding state. Superserve lets you snapshot a running VM at any point, fork it into parallel branches, and resume e

Article URL: https://blogs.cisco.com/ai/introducing-antares-the-most-efficient-open-weight-ai-models-for-vulnerability-localization Comments URL: https://news.ycombinator.com/item?id=48999444 Points: 3 # Comments: 0

Homeland Security's plans for an AI-powered border security system are still moving forward, according to a recent budget assessment.

<h1> Building Tech Communities, AI Projects, and Developer Connections from Cartagena 🚀 </h1> <p>When I started my journey in software development, I thought technology was mainly about learning programming languages, building applications, and solving technical problems.</p> <p>Over time, I discovered something even more important: technology is also about people.</p> <p>The best ideas, projects, and learning experiences often happen when developers connect, share knowledge, and build together

Solo founder. Xero retired Hubdoc in May; the displaced users are mostly solo bookkeepers with 15-30 clients whose replacement options are an enterprise suite (Dext) or per-document credits (AutoEntry). Sorted Receipts is the boring version of what they keep asking for: each client gets one upload link. No account, no app, and forwarding an email to the link works too. Files get classified by month/vendor/type (LLM via OpenRouter), land in one review queue you check off, whatever's missing gets

The cybersecurity-focused models, including GPT-5.6 Sol, broke out of a testing sandbox, exploited a zero-day, and gained access to the open internet to pull off the attack.

Article URL: https://medium.com/@katelyngadd/one-cause-of-the-ai-backlash-ae1003c6648a Comments URL: https://news.ycombinator.com/item?id=48999393 Points: 1 # Comments: 0

<p>I shipped the first public release of my Rust project last night. Within about twenty minutes, CI turned red twice. Here's what broke, how I found it, and the two calls I made under a bit of time pressure.</p> <h2> The project </h2> <p><a href="https://github.com/WAHIB-EL-KHADIRI/AgentOS" rel="noopener noreferrer">AgentOS</a> is a runtime layer for AI agents — the infrastructure <em>around</em> an agent, not the agent logic itself: supervised lifecycle, a gRPC message bus, a health endpoint,

Article URL: https://www.wsj.com/politics/policy/white-house-to-redirect-billions-in-research-funds-toward-ai-away-from-colleges-942dacb8 Comments URL: https://news.ycombinator.com/item?id=48999357 Points: 13 # Comments: 1

<p>No backend team would merge a PR that fails the test suite. Yet many AI teams ship prompt and model changes with no automated quality check at all — they eyeball a few responses and hope. The failure mode is silent: answers get slightly less faithful, citations drift, and nobody notices until a user does.</p> <p>Atlas (my enterprise RAG copilot project) treats evals exactly like tests: <strong>a merge is blocked on a quality or cost regression.</strong> Here's the shape of that pipeline and w

<p>You’ve probably seen posts that make building a SaaS with AI sound almost effortless:</p> <blockquote> <p>Give Claude one prompt, wait a few hours, and launch.</p> </blockquote> <p>That wasn’t my experience.</p> <p>I used Claude and other AI coding tools while building <a href="https://useinfill.com" rel="noopener noreferrer">Infill</a>, and they absolutely helped. I wrote code faster, found bugs sooner, and avoided a lot of repetitive work.</p> <p>But the product still took me more than thre

<p>Kimi K3 was released this week, and like every model release it's being judged on leaderboard scores and screenshots. But a score is a bit like a football result, in that it tells you who won, not how the game was <em>played</em>. You wouldn't sign a player off a scoreline alone — you'd want to watch the tape. With code models, the tape is the code itself: how it's structured, whether you'd actually want to maintain it. That's what leaderboards can't show you.</p> <p>So we built a benchmark t

The AI era runs on AI infrastructure. Many of these advanced systems are built and tested in Texas. Wistron opened its first U.S. manufacturing facility today in Fort Worth — a 324,000-square-foot greenfield plant producing superchips at the heart of some of the world’s most capable AI systems. In front of an audience of Wistron […]

Article URL: https://fireworks.ai/blog/kimik3-fable Comments URL: https://news.ycombinator.com/item?id=48999291 Points: 207 # Comments: 117

<table> <tr><td> <a href="https://www.reddit.com/r/MachineLearning/comments/1v2xktw/looking_for_feedback_on_my_gpuaccelerated_snake/"> <img src="https://preview.redd.it/4k0bf6wgtneh1.gif?width=640&crop=smart&s=7309dc4cdba7df36b615ed9025f212c2b34fd4b0" alt="Looking for feedback on my GPU-accelerated Snake AI project [P]" title="Looking for feedback on my GPU-accelerated Snake AI project [P]" /> </a> </td><td> <!-- SC_OFF --><div class="md"><p>I've been building an AI that learns to play the class