Every story we have published on AI: agents, models, chips, funding, and the race to build what comes next.
Artificial intelligence is no longer a research curiosity. It is the defining technology story of the decade, reshaping how software is built, how companies operate, and how billions of people work and communicate.
This hub collects all of our AI coverage in one place: from the quiet revolution in small, efficient models to the $150 billion funding frenzy, from GPT-6 building you tools on demand to the autonomous agents now doing real jobs at real companies.
We cover AI as practitioners, not spectators. Bookmark this page; it updates every time we publish a new AI story.
Autonomous AI agents are now doing real jobs at real companies: closing books, triaging support tickets, and writing production code. The enterprise pilot era is over.
The AI funding boom is real, but it's a barbell: mega-rounds for proven winners, scraps for everyone else. Inside the most lopsided venture market in history.
On October 6, OpenAI dropped 722 AI-generated mathematics manuscripts, claiming progress on the Riemann hypothesis and more. The math world split instantly, and nobody has verified a single proof yet.
TSMC's third-quarter revenue surged to a record NT$1.49 trillion on relentless AI chip demand. With Foxconn up 47% and banks syndicating a $60 billion chip-financing package, the AI infrastructure boom shows no signs of cooling.
While the industry chases scale, the most interesting breakthroughs are happening in models small enough to run on your laptop.
OpenAI's GPT-6 turns ChatGPT from a text chatbot into an interface generator that builds calculators, diagrams, and interactive tools on demand. With 1.2 billion weekly users, this is the biggest UI shift in consumer software history.
Microsoft's new Surface Laptop Ultra runs large AI models entirely on device, powered by Nvidia silicon. At $2,599, it's a bet that local AI is worth double the price of a normal laptop.
Manus parent Butterfly Effect raised over $500 million after Beijing ordered Meta to unwind its $2 billion-plus acquisition. Investors are still betting big on independent AI companies.
Researchers keep hitting the same wall. Teaching a neural network something new still risks erasing what it already knew. Here's what's actually new.
OpenAI is testing visual ads inside ChatGPT image generation as weekly users hit 1.2 billion. The shift from subscriptions to advertising is the biggest business model bet in AI, and the FTC is watching.
Personalized AI tutoring for every student sounds like a dream. The reality in schools is more complicated, and more interesting.
A chatbot answers one question at a time. An agent plans and executes multi-step work: it uses tools, calls APIs, writes code, and completes tasks across apps like email and calendars. The 2026 enterprise agent wave, including Google's universal work agent and agentic collaboration suites, treats agents as coworkers with their own inboxes and spending limits, not just chat windows.
Roughly $150 billion flowed into AI startups in 2026, but most of it went to about 20 companies with proven products and access to compute. Training frontier models costs hundreds of millions of dollars, so investors piled into the winners. The result is a barbell market: mega-rounds at the top, scraps for everyone else.
Both. Frontier labs keep scaling training runs, while a parallel revolution is making small models dramatically more capable. Efficient models now run on laptops and phones, handle coding and math, and cost a fraction to serve. That is why device makers like Microsoft are betting that local AI is worth a premium.
The problem is called catastrophic forgetting: retraining a neural network on new data can overwrite what it already knew. Continual learning research aims to let models accumulate knowledge the way people do, but as of 2026 it remains one of the field's hardest unsolved problems.