AI News Today

<!-- SC_OFF --><div class="md"><p>I am doing some work with cell type classification, where I have 4.3 million cells and 512 features (condensed embeddings from the encoder of a transformer).</p> <p>The broader goal is to implement a contextual bandit for augmenting the training set of the dataset, as it is currently imbalanced, and rare cell type classification is poor when I tried a baseline logistic regression classifier.</p> <p>Dataset:<br/> Feature matrix shape: (4290471, 512)<br/> Labels s

I was checking benchmarks of muse spark and first it looked impressive that it scores 80 on terminal bench 2.1, but then i saw vals ai's post: in which they are showing that muse spark scores 69.29, which again proved my hypothesis true that benchmakrs are really inconsitent and does not tell whole story. how to trust these benchmarks? what would you say? or i am missing smth? Comments URL: https://news.ycombinator.com/item?id=48854838 Points: 1 # Comments: 0

<p>Dans la plupart des entreprises que je côtoie en Afrique francophone, la gestion RH tient encore sur des fichiers Excel, des cahiers de présence et des groupes WhatsApp. Résultat : des retards non suivis, des bulletins de paie faits à la main, et zéro visibilité pour la direction.</p> <p>J'ai construit <strong><a href="https://eaglecheck.saeicubetech.com" rel="noopener noreferrer">EagleCheck</a></strong> pour changer ça.</p> <h2> Le problème </h2> <p>Les SIRH classiques (Systèmes d'Informatio

Article URL: https://www.ft.com/content/753bd818-524a-46f3-9026-e2fae74c2167 Comments URL: https://news.ycombinator.com/item?id=48854723 Points: 2 # Comments: 0

<pre style='white-space:pre-wrap;width:81ex'>test(agents): stabilize MCP timeout retry coverage (#103223)</pre>

<p>A property inquiry has a half-life measured in minutes. Someone is scrolling listings at 11pm, they fill out the "request more info" form on the three-bedroom on Maple, and they fire the same form off to four other listings on the same street. Whichever agent replies first — with answers, not a "thanks, I'll be in touch" — is the one who gets the showing. By the time a human checks email at 9am, that lead has already toured two other houses in their head.</p> <p>Most "AI for real estate" tool

<p>SDR automation dies when replies hit a no-reply box. The sequence engine is the easy part — anyone can fire "step 1 of 5" on a cron. The part that quietly breaks every outbound tool is the <em>other</em> direction: a prospect reads your third touch, types "not now, circle back in Q3," and hits reply. That reply is the single most valuable signal in the whole campaign, and it lands in <code>no-reply@yourcompany.com</code>, which is to say it lands nowhere. The prospect who told you exactly whe

<p>Law-firm intake is one of those problems that looks like a chatbot and isn't. A prospective client emails <code>intake@yourfirm.com</code> describing a dispute, and what the firm actually needs is careful <em>capture</em> and a clean <em>handoff</em> — collect the matter details, check the names against existing clients and known adverse parties, and put the whole thing in front of a human attorney. The one thing you must not do is let a model freelance legal advice over email.</p> <p>Most "A

<p>Most "AI for collections" demos are a glorified mail merge: a model writes three increasingly stern paragraphs, a cron job fires them on days 7, 14, and 30, and everyone claps. That's fine right up until a customer replies "we paid this last Tuesday, check your records" — and your one-way email service can't read the reply, so the agent sends the day-30 nastygram anyway. Now you've insulted a paying customer and your AR team is on the phone apologizing.</p> <p>Collections isn't a send-only pr

<pre style='white-space:pre-wrap;width:81ex'>feat(webchat): dreams scene sleeps your seeded lobster cameo (#103167) The dreams page's generic sleeper is replaced by the agent's own seeded lobster (same look system as the sidebar pet) rendered with a new sleeping option: pupils hidden, closed-lid curves inline-visible so the cameo needs no lobster-pet CSS context. Palette vars are supplied by the cameo wrapper. Deletes the bespoke dreams svg.</pre>

Would you trust an AI to assist you with financial decision making? What would be some requirements to be comfortable with it? Comments URL: https://news.ycombinator.com/item?id=48854585 Points: 1 # Comments: 1

Article URL: https://selvedge.sh/ Comments URL: https://news.ycombinator.com/item?id=48854574 Points: 1 # Comments: 0

<blockquote cite="https://help.openai.com/en/articles/20001275-chatgpt-work-and-codex"><p>[...] Work on web and mobile runs in the cloud. Work in the desktop app can also use local files and desktop apps with your permission. At launch, cloud Work conversations do not appear in desktop Work; desktop Work threads and local files remain on that computer.</p></blockquote> <p class="cite">— <a href="https://help.openai.com/en/articles/20001275-chatgpt-work-and-codex">OpenAI</a>, trying (unsuccessful

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

i personally believe that the internet is the best thing we've ever built, and arguably, nothing we made before comes close. now agents can use it too, but the web treats them as second-class. every agent works for a person, so whatever it keeps from agents it keeps from us. we believe that if we let them in, each of us gets the whole of what we built, a thousand times over. and to add to that, democratization of tools that allow this is essential, despite the fact that there will always be bad

<p>Quick confession: this started as "let me stop babysitting my tests and just make them stay green," and it turned into something a lot bigger. I want to be upfront about where I'm actually trying to take it.</p> <p>I built Claude Basecamp, and as of today it's open source. The reconciliation loop (declare "tests always green," it holds that true) is the part you'll notice first, but it's not really the point. The point is I want this to become the operating system for everything you do with C

Article URL: https://www.anthropic.com/news/ust-claude Comments URL: https://news.ycombinator.com/item?id=48854398 Points: 2 # Comments: 0

<p>Plans to build a NZ$3.5bn datacentre in Makarewa in the country’s south has drawn concern about electricity and water use, and potential noise pollution</p><p>People living near the site of New Zealand’s first planned AI datacentre are calling for more transparency about the project, especially about how the centre’s huge electricity and water use and potential noise pollution could affect them.</p><p>Singapore-based company Datagrid has secured approval to build a NZ$3.5bn (US$2bn) AI datace

<p> AI agents that verify employment by phone, email & fax </p> <p> <a href="https://www.producthunt.com/products/superunit?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/1192497?app_id=339">Link</a> </p>

Article URL: https://apps.apple.com/us/app/lea-by-legali-ai/id6759547127?see-all=reviews&platform=iphone Comments URL: https://news.ycombinator.com/item?id=48854316 Points: 1 # Comments: 1

GPT-5.6 impresses the critics, but Fidji Simo's exit leaves OpenAI's focus — and its org chart — in flux. PLUS: Meta plays catch-up, and the "AI 2027" authors present "AI 2040."

Article URL: https://tryfraym.com Comments URL: https://news.ycombinator.com/item?id=48854222 Points: 1 # Comments: 0

OpenAI's new family of models will continue to power Microsoft's suite of workplace and productivity apps.

Article URL: https://www.cnbc.com/2026/06/30/cleveland-fed-president-hammack-sees-ai-fueling-inflation-says-rate-hikes-may-be-necessary.html Comments URL: https://news.ycombinator.com/item?id=48854188 Points: 3 # Comments: 0

This paper was accepted at the AI4TCI (Workshop on AI for Secure and Trustworthy Critical Infrastructure Systems) Workshop at the International Conference on Availability, Reliability and Security (ARES) 2026. Autonomous negotiation agents are increasingly deployed in high-stakes settings such as insurance and procurement. While cryptographic techniques protect explicitly disclosed constraint values, they fail to address a subtler threat: behavioral privacy leakage, where an adversary infers pri

Learn how to use ChatGPT, start your first conversation, and discover simple ways to write, brainstorm, and solve problems with AI.

Article URL: https://www.france24.com/en/europe/20260601-ukraine-kamikaze-drone-partially-operated-ai-attacking-russian-convoys Comments URL: https://news.ycombinator.com/item?id=48854108 Points: 5 # Comments: 0

Article URL: https://tilion.dev Comments URL: https://news.ycombinator.com/item?id=48854082 Points: 6 # Comments: 0

There is a lot of buzz around this topic but i am really curious about what's actually useful day to day when running AI/LLMs locally on your PC. A few things I'm trying to understand: Which models are you running locally and on what hardware? What use cases have actually worked out considering the speed and memory limitation? At what point did local start making more sense than just hitting an API? What's still not there yet? Feels like we're at an interesting inflection point but hard to cut t

Article URL: https://techstrong.ai/articles/the-openclaw-foundation-reining-in-a-viral-ai-agent/ Comments URL: https://news.ycombinator.com/item?id=48854021 Points: 2 # Comments: 0

<p>Instagram users should check privacy settings after rollout of new Meta AI image generator, advocates warn</p><p>Meta has sparked blowback from privacy advocates for allowing its new AI image maker to generate photos of users with public profiles by default.</p><p>Users of <a href="https://about.fb.com/news/2026/07/introducing-muse-image-meta-ai/">Meta’s Muse Image AI </a>tool, released Tuesday, can tag public Instagram profiles and generate pictures that pull from faces of people featured in

OpenAI's No. 2 executive, Fidji Simo, is stepping down from her full-time role after her medical leave proved longer than expected — a leadership vacuum that comes at a tricky time as the company eyes a possible IPO and races to catch Anthropic in the enterprise market.

OpenAI's Fidji Simo is departing her full-time role as the company's AGI chief and is transitioning to being a "part-time advisor," she said on X. The news follows Simo's original announcement in April that she would take a few weeks of medical leave due to a neuroimmune condition, shortly after she had taken on the […]

Comments URL: https://news.ycombinator.com/item?id=48853669 Points: 5 # Comments: 8

The move comes after Simo took significant medical leave. She will stay on as a part-time adviser.

<p>Newly minted rich and those anticipating huge IPOs are fueling buying and charter spree in the private jet sector</p><ul><li><p><a href="https://www.theguardian.com/news/2026/feb/17/sign-up-for-the-breaking-news-us-email-to-get-newsletter-alerts-direct-to-your-inbox?utm_medium=ACQUISITIONS_STANDFIRST&utm_campaign=BN22326&utm_content=signup&utm_term=standfirst&utm_source=GUARDIAN_WEB">Sign up for the Breaking News US newsletter email</a><em> </em></p></li></ul><p>Aviation lawyer Amanda Applega

Article URL: https://fulcrum.inc/2026/07/09/fable-cifar-speedrun.html Comments URL: https://news.ycombinator.com/item?id=48853424 Points: 4 # Comments: 0

Article URL: https://www.normaltech.ai/p/up-the-stack-how-ais-escape-from Comments URL: https://news.ycombinator.com/item?id=48853327 Points: 3 # Comments: 0

Article URL: https://github.com/backnotprop/orchestrator Comments URL: https://news.ycombinator.com/item?id=48853202 Points: 6 # Comments: 0

Meta Superintelligence Labs released Muse Spark 1.1 on July 9, 2026, alongside a public preview of the Meta Model API. It is a multimodal reasoning model built for agentic tasks, with a 1,000,000-token context window the model actively compacts, zero-shot generalization to new tools and MCP servers, and multi-agent delegation across parallel subagents. Meta's own launch table shows it leading tool use while trailing Opus 4.8 and GPT-5.5 on coding. The post Meta Superintelligence Labs Releases Mu