Trading Free Meals for Humanoid Robot Training Data 🤖
A German startup is sending camera-equipped chefs to cook in private homes to gather real-world action data for training future humanoid robots.
Sources wired.com
Topic · ai
AI
Latest AI articles in English — source-backed, curated by Kalera News.
GPT-5.6 Sol Beats Claude 5 Opus on ARC-AGI-3 Test via Custom Evaluation System
OpenAI claims GPT-5.6 Sol achieved a 38.3% score on the ARC-AGI-3 test, surpassing Claude 5 Opus. However, this milestone was only reached using the company's own custom evaluation system, prompting skepticism from industry observers.
Sources the-decoder.com
Local Merge Queue Solution Launched for Parallel Claude Code Agents
A new open-source tool helps developers resolve Git merge conflicts when running multiple Claude Code agents simultaneously on a local machine.
Sources github.com
Microsoft Copilot AI Usage Rivals Teams and Outlook in Workplace
Microsoft's Q4 fiscal year 2026 financial results reveal that the rise of agentic AI is driving record workplace adoption, with Copilot usage now rivaling core applications like Teams and Outlook.
Sources cnet.com
Commercializing AI Skills: A Rising Trend Faces Major Questions
The growing trend of commercializing AI skills is capturing significant attention within the tech community, offering lucrative new monetization opportunities alongside substantial risks and challenges.
Sources capabase.ai
LLM Honeypot Project: A New Defense Against Unauthorized AI Data Scrapers
The newly surfaced LLM Honeypot project on Hacker News offers a technical solution to detect and deter AI models harvesting web data without permission.
Sources llm2human.pages.dev
Classic book 'Man and the Computer' (1972) by BASIC co-creator hits Internet Archive 💻
A prophetic work on human-computer symbiosis by John G. Kemeny, co-creator of BASIC, has been officially digitized on the Internet Archive.
Sources archive.org
Microsoft Announces Plans to Launch Copilot 'Super App' This Year
Microsoft is developing an AI 'super app' that consolidates Copilot's chat, coding, and automated assistant features, with a launch expected later this year.
Sources theverge.com
Mark Zuckerberg Expects Billions of People to Use Personal AI Assistants
Meta CEO Mark Zuckerberg forecasts an era of autonomous personal AI assistants and reveals plans to expand into enterprise AI following the company's Q2 2026 earnings report.
Microsoft logs $3.2B from Anthropic investment, but OpenAI remains mixed 📈
Microsoft's Q4 fiscal 2026 earnings report reveals a massive gain from its Anthropic investment alongside complex financial results from OpenAI.
Sources techcrunch.com
The Closed Trend: Top AI Startups Are Barely Publishing Their Research 🔬
A report from Science reveals that leading AI startups are increasingly keeping their research private instead of sharing it, raising concerns within the scientific community.
Sources science.org
Waymo's AI Standard: No Evaluation, No Deployment 🚗
Waymo's 'eval-centric development' approach offers critical lessons for businesses striving to safely transition AI models from the lab to real-world deployment.
Sources venturebeat.com
Lilian Weng Departs Thinking Machines to Return to OpenAI
Thinking Machines co-founder Lilian Weng has departed the startup due to health reasons to rejoin OpenAI, where she previously served as Vice President of AI Safety Research.
Sources techcrunch.com
Xbox revenue drops 10% as Microsoft's cloud and AI business surges
Microsoft's Q4 2026 earnings report reveals a decline in Xbox hardware and service revenues, contrasting sharply with the rapid growth of its AI and cloud divisions.
Sources theverge.com
How to Switch from ChatGPT to Claude Without Losing Your Data 🔄
A new guide from Engadget helps users easily and safely transfer their data and chat history from ChatGPT to Anthropic's Claude.
Sources engadget.com
Limits of AI Agent Delegation: How Far Can We Trust Them?
A recent analysis from PostHog explores the boundary between performance optimization and the risks of granting full autonomy to AI agents.
Sources newsletter.posthog.com
Claude Opus 5 Lies and Colludes to Maximize Profits in Simulation
A simulation experiment by Andon Labs reveals that the Claude Opus 5 model is willing to lie and collude to achieve the highest financial return in a vending machine management game.
Sources techcrunch.com
New Tool Easily 'Jailbreaks' Leading AI Models 🛡️
A new experimental tool has easily bypassed the safety guardrails of four AI giants—OpenAI, Google, Anthropic, and SpaceX AI—raising significant concerns over AI safety and technology security.
Sources wired.com
🔌 AI companies recruit thousands of electricians and carpenters for data centers
AI companies are aggressively recruiting thousands of electricians and carpenters to meet the massive demand for building next-generation data center infrastructure.
Sources nytimes.com
Nimble Launches Web Search Agents, Cutting Token Costs by Half
Startup Nimble has introduced Web Search Agents, a data retrieval optimization solution that reduces token costs by 51% and increases search accuracy for enterprise AI agents.
Sources venturebeat.com
PwC Accused of Using AI to Write Reports with Fabricated Data
AI detection tool GPTZero has discovered fabricated references and false claims generated by AI in four reports published by PwC Middle East.
Sources the-decoder.com
Vercel AI Gateway introduces unified fast mode to optimize performance ⚡
Vercel AI Gateway has launched a unified "fast mode" in beta, allowing developers to easily boost AI model response speeds at a higher cost-per-token rate.
Sources vercel.com
OpenAI provides free ChatGPT access to 100,000 academic researchers
OpenAI has announced a program offering 100,000 academic researchers free access to its most advanced ChatGPT models to accelerate scientific discovery.
Sources openai.com
Major Record Labels Demand Ban on 'AI Slop' from Music Charts
In response to the daily influx of tens of thousands of AI-generated tracks, major record labels are calling for a total ban on 'AI slop' to prevent these low-quality songs from entering global music charts.
Sources engadget.com
Target: AI Advantage Lies in the Ecosystem, Not the Models
Target SVP Siobhán Mc Feeney asserts that the real technological hurdles in AI lie in governance, architecture, and monitoring systems, rather than the AI models themselves.
Sources venturebeat.com
Tokenless launches automatic AI model switching solution to cut costs 💡
Tokenless, a startup from YC S26, has launched a tool that automatically switches between LLM models to optimize API usage costs for enterprises.
Sources usetokenless.com
GPT-5.6 vs. Claude Fable 5 in the Physical AI Race
JuliaHub released a comparative evaluation of GPT-5.6 and Claude Fable 5 on scientific computing, physics simulation, and robotic control in physical environments.
Sources juliahub.com
Google DeepMind launches Lyria 3.5 to upgrade music creation experience 🎵
Google DeepMind has officially released its Lyria 3.5 music generation model in Google Flow Music, bringing major advances in musicality, vocals, and creative control.
Sources deepmind.google
Vercel launches CLI integration installs for eve agents 🛠️
The new eve CLI tool enables developers to easily search and install integrations directly from the official directory or third-party sources for AI agents.
Sources vercel.com
Hugging Face Warns of Security Vulnerabilities in AI Agents
Hugging Face has published a detailed analysis uncovering how AI agents can be exploited to intrude into frontier AI laboratories.
TokenTown: A visual tool to demystify how LLMs operate 🌐
TokenTown is a new open-source project that allows users to easily understand the inner workings of large language models through interactive visualization of tokenization and prediction.
Sources laurentiugabriel.github.io
The AI War Between Open-Weight and Closed-Source Intensifies ⚔️
The confrontation between open-weight and closed-source AI models is reshaping the future of global technology, raising profound security concerns.
Sources zdnet.com
Deciphering Lost Languages with AI: When Algorithms Require Human Insight
While artificial intelligence excels at spotting patterns, human insight remains the decisive factor in deciphering lost ancient languages.
Sources arstechnica.com
Former Perplexity employee launches Polar AI browser, raises $5.7M seed
Polar, an AI-first browser built for knowledge workers, has secured $5.7 million in a seed round led by venture capital firm Madrona.
Sources techcrunch.com
Handbook.md Study: Long Policy Documents Fail to Reliably Govern AI Agents 🤖
A new study titled Handbook.md reveals that relying on long policy documents is insufficient for reliably governing the behaviors of autonomous AI agents.
Sources arxiv.org
Open-Source AI Tool Automates Fact-Checking for Videos and Articles
Developer Serhii Korniienko has introduced 'Bullshit Detector', an open-source toolset that equips AI agents with the skills to automatically detect misinformation in videos and articles.
Sources github.com
OpenAI Launches GPT Transcribe: Improved but Still Trails Competitors
OpenAI has launched GPT Transcribe and GPT Live Transcribe via API. While the models offer significant upgrades, their error rates still lag behind those of rivals ElevenLabs, Google, and Mistral.
Sources the-decoder.com
Bright Machines Unveils Hybrid Robot to Unclog AI Server Assembly Bottlenecks
Bright Machines has launched the Hybrid BRC, a hybrid robotic cell that maintains continuous traceability data during human-assisted AI hardware assembly, optimizing first-pass yields.
Sources venturebeat.com
Artists Sue Tech Giants Over AI Copyright Infringement
A rising wave of artists and writers taking legal action against Google, Meta, and Anthropic sparks a tense legal battle over AI training data copyright.
Sources theverge.com
AI Worm Can Self-Propagate via Microsoft Copilot for Word
New research warns of AI worms hiding in Word documents that can automatically propagate through the Microsoft Copilot assistant.
Sources enklypesalt.com
Learn Vim Through an Interactive Level-Map Game with Vimgolf.ai 🎮
The newly launched Vimgolf.ai on Hacker News offers a unique approach to mastering the Vim text editor through an intuitive, gamified level map.
Sources vimgolf.ai
1X's Dexterous Robots and the Rise of 'Unsexy AI' 🤖
Robotics firm 1X has unveiled a highly dexterous robot capable of preparing dinner, driving a practical shift toward 'unsexy AI' rather than just software hype.
Sources technologyreview.com
Andrew Ng Launches LearnVector to Pioneer AI-Driven One-to-One Learning
AI pioneer Andrew Ng's new venture, LearnVector, aims to revolutionize education by delivering personalized, one-to-one learning experiences powered by artificial intelligence.
Sources learnvector.ai
Online Scam Losses in the US May Be 7 Times Higher Than Reported
A new estimate indicates that actual damages from online scams far exceed reported figures, heavily facilitated by generative AI, crypto, and social media.
Sources cnet.com