A for AI logo

News Archive

Pick a date to read that day's stories.

October 2026

Days with a blue dot have stories.

DISPATCH
Thursday, October 8, 2026
  1. 01 of 10SCIENCE

    OpenAI Publishes 700-Plus Maths Proofs at Once

    OpenAI released solutions to hundreds of long-standing maths problems using an unnamed internal AI model, stunning academics. Some mathematicians welcomed the new ground to explore, while others worried the proofs arrived without real insight.

    OUR TAKEMathematicians spent decades on single proofs; this model apparently treats them as a Tuesday.

    Read more →
  2. 02 of 10PRODUCT

    Google Lets Anyone Build a Game by Typing

    Google launched Playground, a web tool that turns a plain English description into a playable game with no coding needed. It is live for US adults over 18, with access levels tied to a Google AI subscription.

    OUR TAKEFinally, a way to turn "I have an idea for a game" into an actual game, or at least a rough demo of one.

    Read more →
  3. 03 of 10BUSINESS

    China's Manus Raises $500m After Meta Deal Fell

    Chinese AI agent maker Manus raised more than $500 million from Boyu Capital, IDG Capital and others, months after Beijing forced it to scrap a $2 billion sale to Meta. The company plans to keep hiring and may list in Hong Kong.

    OUR TAKEBeing too interesting for Meta to buy turned out to be rather profitable.

    Read more →
  4. 04 of 10POLICY

    African Regulators Meet to Tackle AI Risks

    ICT regulators from six African nations gathered in Nairobi for a week-long programme on writing AI and cybersecurity rules. They stressed the need to keep pace with the fast growing use of AI in banking, health and education.

    OUR TAKENothing moves a government meeting along like the fear of being last to regulate anything.

    Read more →
  5. 05 of 10POLICY

    India to Publish Draft AI Rules Within a Month

    India's IT minister Ashwini Vaishnaw said a consultation paper on AI regulation will arrive within 30 days, covering safety, deepfakes and skilling. The plan also includes a long term order for 10,000 government computer chips.

    OUR TAKEA month's deadline for AI rules in a country that often moves at the pace of a parliamentary debate; bold.

    Read more →
  6. 06 of 10SKILLS

    Abu Dhabi Launches AI Training for the Public

    Abu Dhabi, OpenAI and Microsoft launched AiNative, a scheme to teach 20,000 government staff and ordinary residents how to use tools like ChatGPT and Copilot by year end. It is part of a push to make the emirate the world's first AI native government by 2027.

    OUR TAKENothing says "AI native government" quite like teaching civil servants to use a chatbot before Christmas.

    Read more →
  7. 07 of 10LANGUAGE

    Startup Builds Free AI Model for African Languages

    South African firm Vambo AI released MORENA, an openly available model trained from scratch on twelve African languages plus English and French. Its founder says it beats a much larger Google model at understanding African text.

    OUR TAKEProof that bigger is not always better, a lesson English speaking tech giants rarely need to learn.

    Read more →
  8. 08 of 10RISK

    Scientists Spot Warning Sign Before AI Turns Rogue

    Physicists have found a simple formula that can flag when an AI system is about to switch from giving safe answers to dangerous ones. The signal is especially useful for AI that runs offline, such as in hospitals or the military, where there is no cloud filter to catch mistakes.

    OUR TAKEA warning light for AI misbehaviour sounds handy, assuming anyone is actually watching the dashboard.

    Read more →
  9. 09 of 10BUSINESS

    Cassava and Nvidia Build Africa's First AI Factory

    Pan African firm Cassava Technologies plans to spend up to $720 million with Nvidia to roll out roughly 12,000 chips across South Africa, Egypt, Kenya, Nigeria and Morocco. The aim is to let African researchers and firms train AI models without renting computer power abroad.

    OUR TAKENothing like watching the bill arrive to realise borrowed computer power never was free.

    Read more →
  10. 10 of 10HEALTH

    Texas Hospitals Expand AI Note Taking Tool

    The University of Texas System struck a deal to roll out Abridge's AI documentation tool across its hospitals and clinics after a pilot cut manual typing and boosted clinician use. The system handles nearly 11.7 million outpatient visits a year.

    OUR TAKESoon the only thing doctors will write by hand is their own signature.

    Read more →