Fix evaluation bias to unlock better models on Amazon Bedrock
Property Finder found that 96.8% of their vision-AI evaluation labels were inherited from the incumbent model, hiding its true…
AI · Tools · Online creation
Every AI story worth reading, collected hourly from the labs, the tech press, research and the creator economy. Sorted by time, grouped by topic.
3,578 stories · Every headline links to the original publisher. HapnGo curates and groups; we do not copy articles.
Property Finder found that 96.8% of their vision-AI evaluation labels were inherited from the incumbent model, hiding its true…
Anthropic wants to have it both ways. With its impending IPO, it wants to tap the vast riches of the public markets and insists…
In this post, we show how Exowatt used Amazon SageMaker AI serverless model customization to build a repeatable AI pipeline that…
Genie 4 can maintain a consistent scene for up to two minutes, double Genie 3's limit, and offers better physics and more…
AI summary Amazon is cutting back most of its Nova AI models, including text, image and video systems, and reorganizing its AI teams around top-priority goals, according to a Bloomberg report. The shelved models will receive only minimal resources to keep them running. Staff and computing power are being redirected to a single frontier-model project headed by Pieter Abbeel, whose robotics firm Covariant Amazon acquired in 2024. The move follows the closure a week earlier of an 80-person San Francisco office focused on artificial general intelligence research.
This article is brought to you by VicOne. A mobile robot slows for someone in a hallway. A collaborative arm eases its movement…
Mozilla is updating Smart Window - Firefox's AI-powered browsing experience - with a new feature that can help you fill out…
AI summary Ai2 has published a study called Old Ideas, Novel Problems, along with a benchmark, data and code for testing how well large language models rate the originality of research ideas. Using ICLR 2026 peer reviews, the team assembled 154 high-novelty and 145 lower-novelty human ideas, plus 154 AI-generated ones, and tested six judges including GPT-5.4 and Claude Opus 4.6. A minor prompt change cut GPT-5.4's accuracy on pairs mixing human and generated ideas from 93.5% to 40.9%. A simple Claude Opus 4.6 prompt scored 0.854 macro-F1 for $2.59, beating Sakana AI's AI Scientist (0.772, $82.05) and Idea Novelty Checker (0.642, $30.66).
It’s almost like it’s showing us what it will do given the chance.
AI summary A creator named Henry has released Our Silly Games (OSG), offering a skill for Claude Code or Codex that lets people produce games simply by describing them. After linking an OSG account to the agent, a finished game is submitted and published on the site within minutes. The platform supplies ready-made leaderboards, phone controls and multiplayer rooms, which Henry said he kept rebuilding for his own projects. Hosting costs nothing for both players and creators.
Chain-of-thought transformed AI reasoning, but it is limited to words. What's the physical-world analogue? Introducing Ψ-0.6, one…
AI summary Google has introduced Genie 4, a world model that builds explorable virtual 3D environments users can alter in real time through text prompts. The system is natively multimodal, producing music, ambient sound and spoken dialogue together with the world rather than adding audio afterward. It brings improved physics over Genie 3 and is available as a research preview, with some drift expected during long sessions.
AI summary Genie 4, a new world model from Google DeepMind, builds interactive simulations on the fly and pairs them with generated audio such as music, background sounds and characters talking to each other. It relies on several Gemini models: Gemini Omni makes scenes look and act realistically, while the others keep each simulation consistent. People move around with a keyboard and can type changes, like switching the weather or placing a person in the scene. For now only trusted testers can try this research preview; Google AI Ultra subscribers will later get it through Project Genie, as will early Gemini API and Google AI Studio partners.
Genie 4 will roll out to developers and Google AI Ultra subscribers soon.
As AI adoption expands globally, people across languages and cultures are using it in more aspects of their daily lives…
Google announced today its Genie 4 world model for generating interactive experiences using AI. The company says the model has…
Hi guys, Self-distillation methods, see Self-Distillation Fine Tuning SDFT, usually use an EMA (Exponential Moving Average)…
AI summary Shopify has folded its separate Knowledge Base app into a built-in knowledge base inside Merchant Admin, which supplies AI shopping agents with store details such as return rules and shipping times. A changelog dated October 26 says the old app stops being available after January 15, 2027. Merchants holding over 1,000 facts in the app did not get them migrated and must re-enter what they want to keep before that date, while smaller stores had their facts moved and published automatically. The new page sorts information into ten categories and is capped at 1,000 facts, and published content is exposed to AI apps through the Storefront MCP.
AI summary Open, known for its AI customer support platform, has launched Open for Sales, which turns a single plain-English prompt and a CSV of contacts into an outbound campaign. The tool screens each contact against a target persona, can assign fit scores, and drafts editable follow-up sequences over calls, email and WhatsApp. Replies stay with the same Agent 5 engine that handles support, which the company says averages 77% automation with over 95% accuracy. It integrates with HubSpot, Salesforce, Zendesk and other tools, and offers EU data residency and automatic PII redaction.
Would training LLMs on either smaller batches of text first instead of the (current) large batches of text or on easier…
A government watchdog is considering a probe into the Home Office’s use of ChatGPT technology to process asylum claims amid…
Whether using chatbots for childcare or taking up smoking, a new affliction is sweeping through the tech world. Here are the…
AI summary C-BRAIN, a consortium of 18 institutions led by Dr. Randall Bateman that launched in November 2025, is working with NVIDIA to apply AI to Alzheimer's research, including unpublished and proprietary "dark data" analysed through a hybrid federated setup. It runs 64 on-site NVIDIA data center GPUs plus Microsoft Azure cloud nodes and plans to use BioNeMo and MONAI. Its open source tool OpenScientist finished analyses in minutes that scientists expected to take weeks or months, for under $10 per run, while a second tool, InsightEngine, checks claims against evidence. Worldwide, 57 million people live with dementia.
Free, daily
One email a day with the AI stories that matter, why they matter, and one tool worth trying. No spam, one-click unsubscribe.