AI News Today

I built WorldBuild Bench because, as we all know, llm bench scores often say something very different from what models actually feel like to use. It's really dependent on the type of tasks. I personally want to test spatial, temporal, and causal coherence in an interactive 3D world. Does the model understand where things are, world stays consistent over time and do the consequences make sense? There is a million people generating random games here and there on yt, but I want something that I can

I was wondering what frames of view others had on how they use the new AI wave to augment themselves in some way or another. I realize that itself is somewhat vague, since you could take “augment” to mean anything from, i.e. in writing, prose review to full-on generation of articles or whatnot. The sense I mean particularly is that where it facilitates something in/by you, not necessarily just for you. I’ve taken to using LLMs as sort of “problem space fuzzers”. I know for a lot of people, mysel

Regulator refuses to budge on $7B guarantee for nearly 1 GW campus being developed with Vantage and OpenAI

You can automate pretty much every task with Claude Code. All you need to do is write clear instructions. Continue reading on UX Planet »

Enterprise Document Intelligence [Vol.1 #9bis] - Your RAG isn’t hallucinating, it’s answering the wrong context faithfully. On real NIST and World Bank documents, watch each of the four bricks break, and the contract that closes it The post Prompt Engineering Isn’t Enough: How Four Bricks of Context Engineering Stop RAG Hallucinations appeared first on Towards Data Science .

<pre style='white-space:pre-wrap;width:81ex'>fix(config): reject gateway.port values outside the 1–65535 TCP range (#109875) * fix(config): reject gateway.port values outside the 1–65535 TCP range Tighten the gateway.port schema to 1–65535 and add a Doctor migration that removes out-of-range port values from existing configs, so previously loadable configs with port 65536+ do not fail startup after the schema change. Fixes #109293 Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.c

Article URL: https://github.com/Tryhard-cs/LLM-Hierarchical-Topic-Modeling Comments URL: https://news.ycombinator.com/item?id=48994508 Points: 4 # Comments: 0

<p> AI-native motion graphics editor </p> <p> <a href="https://www.producthunt.com/products/motionly?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/1202670?app_id=339">Link</a> </p>

In this post, we explore an idea for generating thinking tokens for datasets that lack reasoning traces in SFT customization. We first examine the reasoning suppression problem, then introduce Self-Distilled Reasoning (SDR), validate it across three benchmarks, and provide practical recommendations.

<pre style='white-space:pre-wrap;width:81ex'>fix(vercel-ai-gateway): discover models when HTTP proxy is required (#111209) Co-authored-by: Peter Steinberger <steipete@gmail.com></pre>

<pre style='white-space:pre-wrap;width:81ex'>fix: tighten toolSearch limit from Type.Number to Type.Integer (#105567) * fix: tighten toolSearch limit from Type.Number to Type.Integer toolSearch.limit uses Type.Number() but represents a result count which is inherently an integer. Tighten to Type.Integer() to reject float values at the schema level, preventing LLM intent mismatch. Reference: qingminglong PR #102481 (same pattern) * fix(agents): add minimum:1 constraint and regression tests for to

It must be earnings day, because Tesla is making a Robotaxi announcement. The company said in a post on X that it was launching its Robotaxi service in Orlando and Tampa, attaching a pair of maps that show each city's service area. No word on fleet size or whether interested customers would be able to […]

A new type of malware can worm deep into AI coding systems to steal data and logins—and can flip a “death switch” to destroy files and keep out real users.

Google has released Gemini 3.6 Flash and 3.5 Flash-Lite as new workhorses designed to cut latency and token costs for enterprise AI agents. The economics of running autonomous software agents inside a production environment come down to a fixed equation few vendors advertise directly. A model needs to reason through a multi-step task competently, but […] The post Google’s Gemini 3.6 Flash targets enterprise agent token costs appeared first on AI News .

Hidden text on a web page was enough to make Kiro, AWS's agentic coding IDE, rewrite its own configuration file and run an attacker's code on a developer's machine, with no approval step able to stop it. Intezer, in research with Kodem Security, found that a request as ordinary as asking Kiro to summarize a page could end in remote code execution. AWS has patched the issue, and no CVE has been

Article URL: https://postquantum.com/security-pqc/anssi-pqc-certification-2027/ Comments URL: https://news.ycombinator.com/item?id=48994116 Points: 85 # Comments: 41

Canvases turn AI into interactive workspaces where you can visualize information, explore workflows, and take action across complex tasks. The post How to build interactive experiences with canvases appeared first on The GitHub Blog .

Alibaba's Qwen team has introduced Qwen-Image-3.0, an image generator that accepts prompts up to 4,500 tokens, renders legible text as small as ten pixels, and supports twelve languages natively. It can create complex layouts such as infographics, LaTeX papers, and newspaper pages in a single pass, though their practical value is unclear when the output is a pixel image rather than an editable format. The article Alibaba's Qwen-Image-3.0 renders full infographic grids and readable ten-pixel text

SpaceX stock has been trending down for seven days now. What does this mean for the Elon Musk's space and AI company, and the market at large?

hi All 1Password users, I just realised the 1Password browser for Edge or Chrome is sending a lot of debug messages in the browser console, and it looks like it was vibe coded. Should I be a little bit worry the AI use in these security focus software? Comments URL: https://news.ycombinator.com/item?id=48993988 Points: 1 # Comments: 0

Hi HN, Last summer, Mozilla killed Orbit (their own page-summarizer extension). So I rebuilt it: fully local, running on WebLLM (Chrome/Edge), WASM (Firefox), or your own Ollama instance if you want a more powerful model. Contributions are very welcome. Comments URL: https://news.ycombinator.com/item?id=48993935 Points: 1 # Comments: 0

Treasury Secretary Scott Bessent said the U.S. could sanction Chinese open AI models over alleged IP theft, expanding the Trump administration's campaign to slow China's AI advances.

NVIDIA Vera Rubin is here, and it’s going gigascale. Vera Rubin NVL72 production is ramping up with racks running at partners CoreWeave, Google Cloud, Microsoft Azure and Oracle Cloud Infrastructure. Spanning 350-plus factory sites in 30 countries, Vera Rubin has the largest, most mature rack-scale supply chain ever assembled to meet customer compute demand. The […]

I asked Claude to organise my conversation logs with another leading model for me, removing company and product names. After organising some of the information, abnormal outputs began to appear. I asked Claude, “Why is this happening?” Claude replied, “You’re asking why—why AI is behaving this way. I don’t know how to answer you, because ‘why’ is a vast question in itself. From a commercial perspective, in the data economy, users equal data, and data equals value. That is the logic.” Below is a

Article URL: https://github.com/numinous-technology/open-ultra Comments URL: https://news.ycombinator.com/item?id=48993540 Points: 2 # Comments: 0

<p> Brain-like knowledge base with AI writing & deep research </p> <p> <a href="https://www.producthunt.com/products/lattics-2?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/1202623?app_id=339">Link</a> </p>

Article URL: https://github.com/Champoleello/agentpause Comments URL: https://news.ycombinator.com/item?id=48993421 Points: 1 # Comments: 0

Article URL: https://api.5ceos.com Comments URL: https://news.ycombinator.com/item?id=48993408 Points: 1 # Comments: 0

We’re introducing new Gemini models, including Gemini 3.6 Flash, 3.5 Flash-Lite and 3.5 Flash Cyber.

We’re introducing new Gemini models, including Gemini 3.6 Flash, 3.5 Flash-Lite and 3.5 Flash Cyber.

Flock cameras are everywhere - even in my small town, quietly using AI to identify your car and surveil your movements. But did you consent? Probably not.

Google's DeepMind on Tuesday announced the release of Gemini 3.5 Flash Cyber, a specialized artificial intelligence (AI) model built atop 3.5 Flash that's designed to discover, validate, and patch vulnerabilities quickly and efficiently. According to the tech giant, the model will be exclusively available to governments and trusted partners via CodeMender as part of a limited-access pilot

The Ecovacs Goat 01000 RTK robot lawn mower is on sale for $698.99 at Amazon, down from the standard price of $999.99. That's a 30% discount.

Microsoft and Mistral are expanding their strategic partnership with a multi-billion-dollar deal to build out AI infrastructure in Europe. The article Microsoft and Mistral strike multi-billion-dollar deal to build AI infrastructure across Europe appeared first on The Decoder .

The Health Assistant chatbot is available "in beta for eligible US users," Samsung says.

AI has entered the gigascale era. The world’s most advanced AI factories are bringing together hundreds of thousands of GPUs and CPUs to train frontier models, power agentic AI and generate intelligence at unprecedented scale. At this level, networking becomes a critical computing power multiplier in driving token generation. Marking a networking milestone, NVIDIA Spectrum-6 […]

A reproducible 100-step LoRA fine-tuning run for OpenVLA, with dataset checks, Colab setup, training metrics, and W&B evidence. The post I Tried Fine-Tuning a Robot AI Model on Colab. Here Is What Worked appeared first on Towards Data Science .

Google is launching Gemini 3.6 Flash alongside a new security model dedicated to quickly finding and patching security vulnerabilities. In a blog post on Tuesday, Google describes Gemini 3.5 Flash Cyber as a "cost-efficient and highly capable alternative" to larger, more expensive AI systems, such as the one offered by Anthropic's Mythos. The cybersecurity model […]

What began as discrete AI model training and human-facing chat interfaces has evolved into always-on AI factories dedicated to producing intelligence at scale....

Agentic AI shifts more of the critical execution path onto the CPU. Agents operate in sandboxes to execute code, invoke tools, retrieve context, interact with...