AI News Today

<p>It was 1 a.m. when the PM dropped a screenshot into the group chat: “Your AI assistant forgot the customer’s name again. Third time.” I zoomed in and dragged my finger across the words “Dear user, hello!”. My heart sank — the “long-term memory module” we had just shipped was completely unreliable in the real world. I had manually tested dozens of conversations. It always <em>felt</em> fine, but the moment it hit real users, everything fell apart. I closed Slack, opened my IDE, and decided to

<p>I am a Pakistani software founder who built a $100K MRR AI SaaS over three years. In May 2026, I negotiated a $200K acquisition with <strong>James De Berardine</strong>, Director of NOTO nightclub in Philadelphia.</p> <p>I transferred my domain and full proprietary codebase to his AWS environment in good faith. He made six written promises to fund a $30K deposit. He paid $1,000. He retained my code for two months while demanding endless diligence. When pressed to close, he called the transact

Article URL: https://ghostinterview.dev Comments URL: https://news.ycombinator.com/item?id=48839534 Points: 1 # Comments: 0

<p>GPT-5.6 Sol, Terra, and Luna are becoming the center of the coding-agent conversation this week.</p> <p>The important question is not only whether Sol is better at hard coding tasks, or whether Luna is cheaper for repetitive work. The more useful question is what happens after the agent produces a working app.</p> <p>For many builders, the pattern is already familiar:</p> <ol> <li>A coding agent builds a convincing demo.</li> <li>The local app works.</li> <li>The first real user requirement a

<p><a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fvai0v9eiruy5k8ci6o1s.png" class="article-body-image-wrapper"><img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fvai0v9eiruy5k8ci6o1s.png" alt=" " width="799" height="4

Ant Group's Robbyant has released LingBot-VLA 2.0, an Apache-2.0 vision-language-action model for cross-embodiment robot manipulation. The 6B checkpoint is pretrained on roughly 60,000 hours of data, spanning 50,000 hours of robot trajectories across 20 robot configurations and 10,000 hours of egocentric human video. It maps every embodiment into a single 55-dimensional canonical action space, covering arms, dexterous hands, waists, heads, and mobile bases. A token-level, auxiliary-loss-free Mix

Article URL: https://github.com/tiliondev/fortress/tree/main/mcp Comments URL: https://news.ycombinator.com/item?id=48839328 Points: 4 # Comments: 0

<pre style='white-space:pre-wrap;width:81ex'>test(nvidia): align guided secret metadata</pre>

built it in spring 2026--ready to share today. use opus 4.8 to create css and js snipptes to style the websites you visit the way you want. make nytimes more print-like, hn more matrix, or just replace any mention of AI with the :poop-emoji" -- your call. Comments URL: https://news.ycombinator.com/item?id=48839159 Points: 1 # Comments: 0

Three humanoid companies moved toward the public markets in a single week. Agility filed to go public via SPAC at $2.5 billion, Unitree cleared its Shanghai IPO, and Tesla started turning the line that built its last Model S into an Optimus factory. Mistral shipped a robot brain that finds its way with one cheap camera. And the research this week kept landing on the same catch: locomotion is getting solved, while the models still lose basic world knowledge the moment you train them to act. The m

Multimodal large language models (MLLMs) have recently shown strong performance in visual understanding, yet they often lack temporal awareness, particularly in egocentric settings where reasoning depends on the correct ordering and evolution of events. This deficiency stems in part from training objectives that fail to explicitly reward temporal reasoning and instead rely on frame-level spatial shortcuts. To address this limitation, we propose Temporal Global Policy Optimization (TGPO), a reinf

GPT 5.6 is now available on AI Gateway in three models: Sol, Terra, and Luna. GPT 5.6 from OpenAI is now available on AI Gateway in a limited preview, across three models: Sol, Terra, and Luna. All three are stronger at agentic work across coding, biology, and cybersecurity, and are more token-efficient than the previous generation. Sol ( openai/gpt-5.6-sol ): the flagship, and the most capable of the three. Terra ( openai/gpt-5.6-terra ): a balanced model for everyday work, with performance com

Long-context handling remains a core challenge for language models: even with extended context windows, models often fail to reliably extract, reason over, and use the information across long contexts. Recent works like Recursive Language Models (RLMs) have approached this challenge by agentic way of decomposing long contexts into recursive sub-queries through programmatic interaction at inference. While promising, the success of RLMs critically depends on how these trajectories of context-inter

On-policy distillation offers dense, per-token supervision for training reasoning models; however, it remains unclear under which conditions this signal is beneficial and under which it is detrimental. Which teacher model should be used, and in the case of self-distillation, which specific context should serve as the supervisory signal? Does the optimal choice vary from one token to the next? At present, addressing these questions typically requires costly training runs whose aggregate performan

Muse Spark 1.1 from Meta is now available on AI Gateway . It is a multimodal reasoning model with a 1M token context window built for agentic tasks, accepting text, image, video, PDF, and audio inputs. Muse Spark 1.1 plans and orchestrates work across tools and services, operating as a main agent or as a subagent, and it works with new tools, MCP servers, and custom skills without examples. The model supports parallel tool calling, structured output, and built-in search with citations. To use Mu

<p><strong><a href="https://bun.com/blog/bun-in-rust">Rewriting Bun in Rust</a></strong></p> Jarred Sumner has been promising this blog post (<a href="https://x.com/jarredsumner/status/2053063524826620129">since May 9th</a>) about his Zig to Rust rewrite of Bun for significantly longer than it took him to finish the rewrite.</p> <p>Honestly, it was worth the wait. This is a detailed description of an extremely sophisticated piece of agentic engineering, featuring dynamic workflows, trial runs, a

SpaceXAI released Grok 4.5, a Cursor-trained model for coding, agentic tasks, and knowledge work. It serves at 80 TPS, costs $2/$6 per million tokens, and ranks #1 on Harvey's Legal Agent Benchmark. The post SpaceXAI Releases Grok 4.5, a Cursor-Trained Model for Coding, Agentic Tasks, and Knowledge Work at $2/M Input appeared first on MarkTechPost .

Article URL: https://skillburst.ai/ Comments URL: https://news.ycombinator.com/item?id=48838979 Points: 1 # Comments: 0

Guident CEO Harald Braun breaks down the new AV mandates in California mandates and how they will affect the driverless industry. The post Tickets, geofences, and 1M miles: The new reality of California AV compliance appeared first on The Robot Report .

The Amotions AI Zoom App is now available! Win more business opportunities with real-time AI guidance during your Zoom meetings and calls to help you: Ask better discovery questions Answer technical questions with confidence Handle objections effectively Stay aligned with your sales playbook Win more customer conversations Adding the Amotions app in your Zoom desktop app only takes a minute. You can set up Amotions app to auto-start in every of your meetings! It doesn’t join as another participa

Article URL: https://ainowinstitute.org/publications/friendly-fire-exploit-brief Comments URL: https://news.ycombinator.com/item?id=48838857 Points: 1 # Comments: 0

Article URL: https://www.youtube.com/watch?v=2J2Fb1bBufA Comments URL: https://news.ycombinator.com/item?id=48838844 Points: 3 # Comments: 0

Anyone heard or xsight labs (https://xsightlabs.com/)? They are making networking gear for AI and spacex satellites. I'd be curious if anyone can shine a light on: - Does full switch programmability actually matter in production, or is it one of those things that sounds great in a slide deck but nobody uses? - Is "on-path cores" in a DPU genuinely better than Pensando's P4 approach or NVIDIA's off-path model, or does it converge at scale? - Can you actually swap merchant silicon under SONiC in p

Comments

<p><strong><a href="https://openai.com/index/introducing-gpt-live/">Introducing GPT‑Live</a></strong></p> OpenAI <em>finally</em> upgraded the model used by ChatGPT voice mode!</p> <p>I've had preview access for a few weeks in the iPhone app, and the new model is very impressive. It also has the ability to spin off harder tasks to GPT-5.5:</p> <blockquote> <p>For questions that require web search, deeper reasoning, or more complex work, it delegates to our latest frontier model behind the scenes

With GPT-Live, talking, listening, and formulating answers all happen at once

The median cost of a home in the city is now $1.7m, a record high, according to the latest figures.

Hugging Face LeRobot is an open-source robotics library for training, running, and sharing robot datasets, models, policies, and workflows. The post NVIDIA and Hugging Face bring new models and frameworks to LeRobot appeared first on The Robot Report .

Meta might be the next company to make an always-on AI wearable. The company is working on prototype "super sensing" always-aware smart glasses that could continuously record audio and snap photos "every few seconds," according to the Financial Times. The wearer could then ask Meta AI about the captured audio and images. However, the images […]

A few weeks back as the US export ban news hit, and while I still had access to Fable, I asked it to create a blog from his own POV. Of course I "prompted" this and it would be silly to try to state otherwise, but regardless I think its thinking and thus writing are interesting enough to share. The full contents of this site are written by the LLM. From text to code to svg's. It made several articles, I think they're also a bit short and shallow like the lake's itself is, but I'll leave that opi

The newly renamed SpaceXAI wants you to believe little ol' Grok is all grown up

Article URL: https://www.tryagentscore.com Comments URL: https://news.ycombinator.com/item?id=48838101 Points: 2 # Comments: 1

<p> Where human life runs with AI </p> <p> <a href="https://www.producthunt.com/products/expert-chase-deleted-1107920?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/1191527?app_id=339">Link</a> </p>

Elon Musk's SpaceX released Grok 4.5 on Wednesday, the first artificial intelligence model the company has trained specifically for coding and autonomous agents — and the first tangible product of its $60 billion acquisition of the AI coding startup Cursor, completed just weeks ago. The launch marks a pivotal test of the sprawling, vertically integrated AI empire Musk has assembled over the past six months, and of a strategy that bets developers care less about topping benchmark leaderboards tha

Article URL: https://github.com/Actenon Comments URL: https://news.ycombinator.com/item?id=48837953 Points: 1 # Comments: 0

The National Highway Traffic Safety Administration said emergency scenes are not "edge cases."

<pre style='white-space:pre-wrap;width:81ex'>test(anthropic): cover Sonnet 5 CLI allowlist</pre>

AI cheating leads to "a failed society," professor says.

Comments

Article URL: https://github.com/supernovae-st/nika Comments URL: https://news.ycombinator.com/item?id=48837692 Points: 1 # Comments: 0