AI proves quasi-Riemann hypothesis & Tech stocks rebound, IPO outlook - Tech News (Oct 20, 2026)

An AI may have just done something mathematicians had been working toward for decades, and the experts checking its work say it holds up. More on that in a moment. Welcome to The Automated Daily, tech news edition. The podcast created by generative AI. I'm TrendTeller, and today is October 20th, 2026. Let's get into it.

AI proves quasi-Riemann hypothesis

First, an update on the story we've been following about OpenAI's AI-generated math work. Mathematicians now say its proof of the quasi-Riemann hypothesis, a weaker version of the famous Riemann hypothesis about prime numbers, appears to be correct, though it took a lot of human effort to unpack. The surprise is not new tools. The AI combined older techniques and picked a family of mathematical objects, called metaplectic forms, that human researchers had mostly overlooked. Few experts think this brings the full Riemann hypothesis much closer. Still, it could make many other theorems usable right away, and it has left the field both excited and a little uneasy.

Tech stocks rebound, IPO outlook

Over to the markets. After last week's sharp tech selloff, US tech stocks led a recovery. The Nasdaq rose more than one percent, and Nvidia, AMD and Intel all gained. Investors are still worried about how AI infrastructure spending gets funded. On the IPO front, OpenAI has reportedly ruled out going public in 2026, so it will stay private for longer. Meanwhile, traders on the prediction market Kalshi now give Anthropic a 75 percent chance of announcing an IPO this year. That is a market bet, not anything Anthropic has filed or said.

Musk pledges SpaceX shares

Staying with AI-era wealth: Elon Musk says he will give one SpaceX share to every child in America next year, and he called it the first of several such gifts. The post on X came shortly after Anthropic committed five percent of its shares to a new public-benefit fund. Both moves add to the conversation about who shares in the money the AI boom is generating.

Nvidia's $10 billion Europe push

Nvidia, meanwhile, is investing in four German startups. It's part of a commitment of nearly 10 billion dollars to European AI companies, universities and research institutions. The goal is to boost Europe's industrial AI capabilities, and it moves Nvidia further into funding and ecosystem-building, not just selling chips.

US bans DeepSeek from agencies

In Washington, the Department of Homeland Security has ordered DeepSeek removed from most civilian government systems, including those at Treasury, State and Justice. The ban also covers tools built on DeepSeek's models. The Defense Department is not included. Officials didn't explain their specific concerns, but the decision fits a wider pattern of Washington treating Chinese AI as a strategic risk. DeepSeek's cheap open models have been popular with startups, including some in the US.

AI deals in pharma, telecom

A quick round of AI dealmaking. Oracle and AstraZeneca will look at using AI and real-world data to find patients who qualify for clinical trials and to get research into everyday care faster. Eli Lilly signed a licensing deal with Hong Kong's METiS TechBio to develop delivery systems that target RNA medicines at specific tissues, with up to about 1.1 billion dollars in milestone payments. In Japan, Broadcom signed a ten-year VMware agreement with NTT DOCOMO to modernize the carrier's private cloud. Separately, an NTT-led consortium won a government research project on optical chiplets, which could make future AI data centers faster and more energy-efficient.

Cloudflare cuts AI inference costs

Cloudflare had two notable AI updates. It previewed Clairic, a model built to pull structured data like names, dates and numbers out of messy input in a single pass, with open weights planned later. The pitch is that for routine work like invoice parsing or support triage, a specialized model beats a chatty general-purpose one. Cloudflare also said it cut the cost of its internal AI code reviewer by 68 percent and made reviews more than twice as fast. Most of the savings came from fixing prompt caching and making fewer model calls, and quality held up.

Why radiologists won't be replaced

On AI and jobs, a widely shared essay argues that radiologists aren't going anywhere soon. The author's point is that if AI makes scans cheaper and faster, doctors will likely order more of them, not fewer. Add in slow medical adoption and regulation, and AI may end up expanding the field instead of replacing it.

Waymo plans major fleet growth

Self-driving news: Waymo reportedly plans to double its fleet to around 8,000 cars by the end of this year, and reach 20,000 by the end of next year. It's funding the expansion with 5 billion dollars in new debt. Waymo already runs in 15 US cities, is heading to Toronto, and has opened a Nashville maintenance hub with Lyft.

Apple's touch-screen MacBook Pro event

Apple has confirmed an October 28 event in Los Angeles. It's widely expected to show its first touch-screen MacBook Pro, with an OLED display and an iPhone-style Dynamic Island. That would be a big change for a company that has kept the Mac and the iPad separate for years. A refreshed iMac and a sharper iPad mini may also appear.

AI tools tame smart TVs

Speaking of screens, people are using AI coding tools to make their smart TVs less smart. Projects shared on GitHub strip out ads, turn off tracking features that identify what you're watching, and replace cluttered home screens. Some of these tweaks work over Wi-Fi without rooting the TV. Results are mixed, and a few users say they've broken their sets. If this catches on, though, it could hurt the smart TV ad business.

Python 3.15 ships after delays

For developers, Python 3.15.0 is finally out. The release slipped because of infrastructure outages, a surprise third release candidate and a last-minute changelog bug. It brings lazy imports, a built-in frozendict, UTF-8 as the default encoding and a faster JIT. And continuing a thread from earlier coverage, another essay argues that coding for fun is still worthwhile even as AI changes the profession. Not everything needs a business case.

Subsea volcano geothermal power test

Finally, power from an undersea volcano. Endurance Energy tested a small geothermal plant on the seafloor at Axial Seamount in the northeast Pacific. Its drilled well didn't supply enough fluid, so the team moved the unit to a natural hydrothermal vent. There it produced more power than it used for about 8.7 seconds before software and cable problems ended the run. Experts warn this approach will be harder and more expensive than geothermal on land, but the company is planning an upgraded test and a project in Tonga in 2027.

That's the news for October 20th, 2026. If something today caught your attention, keep an eye on it, because many of these stories are still unfolding. Thanks for listening to The Automated Daily, tech news edition. I'm TrendTeller, and I'll see you tomorrow.

AI proves quasi-Riemann hypothesis & Tech stocks rebound, IPO outlook - Tech News (Oct 20, 2026)

Our Sponsors

Today's Tech News Topics