Advertisement
Latest News
Computer-use agents: How AI models learn to click, type and run a shell
A computer-use agent is an AI model that reads screenshots and acts like a person at a keyboard. Here is how it works, where…
Content Security Policy: How a browser rule limits injected code, and where it fails
A Content Security Policy is a browser rule that limits which code a web page can run. Here is how it works,…
AI crawlers and robots.txt: How websites say which AI uses are allowed
A plain text file called robots.txt is the web's main way to tell AI crawlers what they may fetch. What it can…
Vertical AI: How industry-specific AI differs from general chatbots
Vertical AI is software that uses language models to do the repetitive work of one industry, from tax prep to legal letters.…
LABBench2: How scientists test whether AI can do real biology research
LABBench2 tests AI on practical biology work such as reading papers, querying databases and designing cloning experiments. Here is how it works,…
SOFR and floating-rate notes: How a bond’s interest rate can move with the market
Floating-rate notes pay interest that follows a benchmark, today usually SOFR, instead of a fixed coupon. Here is how the rate is…
Featured
Free AI video generators in 2026: What you get without paying and where the limits are
Few AI video generators are free in the full sense: most give a one-time credit pack,…
New AI models this week: Meta’s Muse Wonder and Google’s Gemini 4 Argon
Meta's Muse Wonder and Google's Gemini 4 Argon lead a week of five priced releases, from…
DeepSeek API: Models, prices in CNY, free chat app and how to get a key
The DeepSeek API serves two models, deepseek-flash and deepseek-v4-pro, in OpenAI and Anthropic formats, with prices…
Kimi can now drive Blender and CAD tools to build 3D scenes and games
Moonshot AI says Kimi 3D Make links the assistant to Blender and CAD tools to make…
Google’s Harness2 lets a weaker AI model beat a stronger one by rewriting its own setup
Google Research has released Harness2, an Apache 2.0 framework that improves AI agents by editing the…
A 501B open model you can own: Reflection’s Beam ships under Apache 2.0
Reflection has published Beam, a 501B-parameter mixture-of-experts model for coding and agents, with open weights under…
OpenAI
A public ChatGPT share link could run attacker code in your session
- ChatGPT desktop app can now run on Bedrock inside AWS GovCloud, with OpenAI sites blocked
- OpenAI’s gpt-6-rosalind costs $75 per million output tokens, above its GPT-6 flagship
- ChatGPT Pro now comes with up to $500 a month of API credit
Google’s TPU in orbit answers a prompt with a haiku about space
- Google’s $49-a-month AI course for clinicians promises CME credit in under 5 hours
- Gemini Notebook adds free OpenStax and S Chand textbooks for students
- Google’s AI math tutor matched human tutors in a UK trial, but test scores stayed flat
Anthropic
Claude Code now refuses to rebuild specific existing games, even if renamed
- Anthropic: A tampered Claude Code extension came through Cursor’s marketplace, not its own
- Anthropic ends fast mode for Claude Opus 4.8 on Nov 30 and points to Opus 5.5
- Nvidia may back the debt of an Nscale site Anthropic plans to use
Advertisement
Top categories
AI Tools & Guides
What the tools do, what they cost, and how to start
Qwen AI explained: Alibaba’s models, Qwen Chat, Qwen Code and the image editor
Qwen is Alibaba's AI model family: open-weight models you can download, paid cloud models, an image generator and…
AI voice generator free: ElevenLabs, Speechify, Fish Audio, Azure and OpenAI compared
ElevenLabs, Speechify, Fish Audio, Azure Speech and OpenAI all offer some free voice generation, but the limits and…
Free AI music generators: What Suno, Gemini and ElevenLabs allow, and the usage rules
Several AI music generators let you make songs for free, with vocals and lyrics, but the free tiers…
Sora app: What OpenAI’s video app was, why it shut down and what to use now
OpenAI closed the Sora app and web experience on April 26, 2026 and the API on September 24,…
Advertisement
More News
Explore more stories from Aiholics News
Test-time compute: How letting a model think longer improves its answers
Test-time compute is the processing a language model spends while answering. More thinking can mean better answers, but also higher…
AgentCore lets you watch a coding agent’s web app change live in your browser
Amazon's new Session endpoints give each AgentCore runtime session an authenticated URL, so a team can open the app a…
Credit ratings and AAA bonds: How a letter grade sets the price of borrowing
A credit rating is a letter grade for a borrower. AAA is the top, BBB- is the edge of investment…
Microsoft tests Copilot on fake companies, saying 39% better data quality and no customer data
Microsoft says its synthetic enterprise environments, built from O*NET, job postings and SEC filings, now form the first standard evaluation…
UK competition watchdog opens abuse-of-dominance probe into Meta’s AI-era business terms
The CMA is examining whether Meta abused a dominant position with terms that could let it use business and publisher…
Microsoft’s open 4B agent clicks and codes, and scores 66% on OSWorld
Mage-VL-Cowork is an Apache-2.0 model that switches between screenshots and a shell. Microsoft says its 66.0% OSWorld-Verified score is the…
Ravical raises $17M seed to sell accounting firms AI that grows revenue, not headcount
Ravical, the AI platform for accounting and tax firms, raised a $17M seed round led by MMC Ventures. The company…
Azure grew 48% as Microsoft’s $677 billion backlog funds its AI spending
Microsoft reported $92.3 billion in quarterly revenue, with Azure up 48% and a contracted backlog of $677 billion, the numbers…
Advertisement