A for AI logo

News Archive

Pick a date to read that day's stories.

September 2026

Days with a blue dot have stories.

DISPATCH
Saturday, September 19, 2026
  1. 01 of 10BIOTECH

    Stanford team builds a virtual biotech company run by 37,000 AI agents

    Stanford Medicine researchers have spun out a software-only drug firm staffed by thousands of AI agents that each play a role in the research pipeline. In tests, the system predicted which drug targets were most likely to succeed in trials and independently designed a lung cancer therapy that a major pharmaceutical company later validated.

    OUR TAKEIt is a striking thought: a whole company on a hard drive, with no payroll, no lab rent and no need for coffee breaks. The real question is whether its best ideas still need humans to run the actual experiments.

    Read more →
  2. 02 of 10HEALTH

    Mithrl raises $20 million to speed up drug discovery with a biomedical world model

    The San Francisco startup, founded after a family experience with kidney disease, has built a platform that combines curated biology with agentic tools to help research teams move from idea to clinical trial faster. Its Series A was led by Obvious Ventures.

    OUR TAKEA world model for biology sounds grand, but the goal here is refreshingly narrow: get medicines into trials sooner. The best health AI usually starts with a personal reason rather than a slide deck.

    Read more →
  3. 03 of 10STARTUPS

    A ChatGPT co-inventor’s new startup outputs decisions, not chat

    TypeSafe AI, founded by former OpenAI researcher Diogo Almeida, has released Jev, a model that does not produce text but instead returns calibrated probability decisions. The approach is designed to stop hallucinations and make software automation cheaper and more reliable.

    OUR TAKEAfter four years of chatbots that sound clever but do little, a model that quietly answers yes or no could be the boring breakthrough developers actually need.

    Read more →
  4. 04 of 10STARTUPS

    Manus is reportedly raising $500 million at a $4 billion valuation

    The Chinese AI agent startup, which went viral last year and later saw a Meta acquisition blocked by Beijing, is now in talks for a major new funding round as it resumes independent operations.

    OUR TAKEFour billion dollars for a company most people know from a single viral demo is a reminder that hype still travels faster than revenue in the agent era.

    Read more →
  5. 05 of 10ROBOTICS

    D-Robotics raises $400 million to power humanoid robots with its own chips

    The Chinese semiconductor startup, spun out of Horizon Robotics, designs the Sunrise family of processors that run inside AI-powered humanoids and industrial robots. The round comes the same week a US ban on new Chinese humanoid robot imports took effect.

    OUR TAKEBuilding robot brains and selling them to everyone else is a shrewd bet: you do not have to win the robot race to profit from the runners.

    Read more →
  6. 06 of 10STARTUPS

    Ex-DeepMind researchers raise up to $700 million for a UK world-model startup

    Emulate, incorporated just a month ago in London, is reportedly in advanced talks for one of the largest seed rounds yet for a company building world models. The founders worked on DeepMind's Genie project, which generates interactive 3D environments from images or text.

    OUR TAKEA seed round that rivals most companies' entire funding history suggests investors are wagering that predicting worlds is the next big thing after predicting words.

    Read more →
  7. 07 of 10POLICY

    Nineteen EU countries back a joint public AI project

    The European Commission has welcomed plans for the first Important Project of Common European Interest in AI, covering the full AI stack from computing to applications. The project is coordinated by Germany and aims to support a shared, sovereign European AI ecosystem.

    OUR TAKESovereign AI sounds like a Brussels buzzword, but at least nineteen governments have agreed on something. The hard part will be turning the paperwork into silicon.

    Read more →
  8. 08 of 10POLICY

    South Korea plans a free, unlimited AI chatbot for every citizen

    President Lee Jae Myung wants every resident to have access to a general-purpose chatbot and a public service agent built on domestic models by the end of the year, part of a broader plan to make South Korea a global AI power.

    OUR TAKEHanding a chatbot to fifty-one million people is the easy bit. Making it useful without becoming a $10 billion political football is where it gets interesting.

    Read more →
  9. 09 of 10POLICY

    Singapore signals tighter AI safety rules as it learns from aviation

    Minister Josephine Teo said trust must keep pace with capability and hinted at new safeguards, using civil aviation regulation as an analogy for how to build public confidence in advanced technology.

    OUR TAKEComparing AI to aviation is a useful reframing: people expect planes to be safe because regulators demand proof, not promises.

    Read more →
  10. 10 of 10STARTUPS

    Comp AI raises $34 million to automate security compliance

    The Miami-based startup, founded in early 2025, helps companies stay on top of compliance and cybersecurity using AI agents. It now serves more than 1,000 customers and says revenue has grown fifteen-fold in a year.

    OUR TAKECompliance is the unglamorous corner of tech where nobody wants to spend time. That is exactly why a tool that does the paperwork is worth funding.

    Read more →