AI breaks Erdős geometry conjecture & Google search goes more AI - Tech News (May 23, 2026)

An AI system may have just outdone Paul Erdős on a geometry challenge that’s been sitting open for about 80 years—and independent mathematicians say the result checks out. Welcome to The Automated Daily, tech news edition. The podcast created by generative AI. I’m TrendTeller, and today is May-23rd-2026. Let’s get into what happened—and why it matters.

AI breaks Erdős geometry conjecture

We’ll start with that headline-grabber from the math world. OpenAI says one of its experimental chatbots has produced a new, better arrangement for the classic “unit-distance problem,” reportedly beating a construction Paul Erdős proposed back in 1946. What makes this more than a flashy claim is that independent mathematicians—people not affiliated with OpenAI—reviewed and verified the result. OpenAI also says the solution involved a long chain of reasoning and tools from algebraic number theory to pick point coordinates that satisfy tight constraints. The company hasn’t named the model, and it hasn’t fully released the full write-up, but if this holds up in wider scrutiny, it’s a notable moment: AI not just assisting with proofs, but plausibly generating a genuinely new mathematical result that experts accept as correct.

Google search goes more AI

Sticking with AI—but shifting from the chalkboard to the browser—Google is redesigning its iconic search box for the AI era. The idea is to make it feel normal to type longer, more conversational questions, and to drop in images, video, and even files to guide what you’re searching for. This comes as Google keeps blending AI summaries—its “AI Overviews”—with traditional link results. The upside is speed and convenience for users who want an immediate answer plus sources. The downside, critics warn, is that the more Google intermediates what you see, the harder it becomes to understand why certain information is being emphasized—and the more damage a confident AI mistake can do at scale. There’s also the business ripple: if the answer is on Google’s page, fewer people click out to publishers, which could reshape how the web gets funded.

Gemini moves into smart homes

Google also has a second AI push underway at home—literally. The company says it’s turning Google Home into a “full-stack AI offering,” combining its home APIs with Gemini-powered features so partners—like internet providers, security firms, and device makers—can build and sell services on top of Google’s platform. Read between the lines and you see a strategic shift: more smart-home innovation pushed to partners, more recurring subscription revenue, and potentially less reason for Google to keep producing as many first-party Nest-style devices. For consumers, this could mean smarter automations and more capable assistants—along with a future where your home’s best features might sit behind a monthly plan.

AI-designed miniproteins target GPCRs

Now to biotech, where AI is showing up in a very different form: proteins designed on purpose, rather than discovered by chance. Researchers at the University of Washington’s Institute for Protein Design, working with startup Skape Bio, report they’ve used AI-driven protein design to create ultra-small “miniproteins” that can switch key cell receptors on or off. These receptors—GPCRs—are among the most important targets in modern medicine, but they’re notoriously tricky because their binding regions can be deep and flexible. The team designed proteins under 100 amino acids that can selectively stabilize a receptor in an active or inactive state, and they tested huge numbers of candidates directly in living human cells. There’s also an early animal result: one designed miniprotein performed similarly to an existing drug, with fewer side effects in a mouse study. The big takeaway isn’t a single new medicine tomorrow—it’s a potential new playbook for targeting a family of receptors that medicine relies on, but often struggles to control precisely.

Stem-cell gut tissue with nerves

Another standout in bioengineering this week: researchers describe a “confined culture system” that turns tiny stem-cell-derived gut spheroids into longer, tube-like gastrointestinal tissue by briefly growing them in a 3D-printed scaffold. That matters because conventional intestinal organoids often stay small and ball-shaped, which limits realism and usefulness for transplantation. In this approach, the engineered grafts could be transplanted earlier and engrafted far more efficiently, then matured into centimeter-scale tissue with more lifelike structure. The surprising detail is that the grafts developed a human-origin enteric nervous system—neurons and support cells—without researchers having to add external nerve precursor cells. And those nerves weren’t just decorative: the tissue showed nerve-dependent contractions similar to adult intestine. If this scales and translates cleanly, it could strengthen disease models and bring engineered gut grafts closer to being a practical therapy for intestinal failure.

Parkinson’s LRRK2 gene-silencing trial

On the neurodegenerative front, early human trial results are out for BIIB094, an experimental therapy aimed at the Parkinson’s-linked LRRK2 gene. In a placebo-controlled phase 1 study, the drug was delivered via lumbar puncture, and the primary goals were safety and whether it hits its target. Reportedly, it was generally well tolerated, and spinal fluid tests showed LRRK2 protein levels dropping—by as much as about 59% in treated participants. Importantly, the reduction showed up not only in people with known LRRK2 variants but also in people without them, which hints at a broader potential use. What it doesn’t show—yet—is whether patients actually do better clinically. That’s the work for larger, longer trials. Still, this is a step toward treatments designed to change disease biology, not just manage symptoms.

GLP-1 brain signaling mapped at last

Related to brain and body health, NIH scientists say they’ve opened up a long-standing black box: what semaglutide is doing inside appetite-related neurons. Using real-time imaging in living mouse brain tissue, they traced weight-loss-relevant signaling to a specific messenger molecule, cAMP, in GLP-1 receptor neurons in an area tied to nausea and appetite control. The intriguing part is variability: some neurons sustained strong signaling while others only spiked briefly, which could help explain why people have very different outcomes on the same GLP-1 drug—and why weight loss can plateau. The team also showed that blocking a cAMP-breaking enzyme, PDE4, pushed more neurons into a longer-lasting response, suggesting a possible future for combination therapies. It’s early-stage biology, but it points to tangible knobs researchers might be able to turn for more durable effects.

Wearable AI uses muscle stimulation

From medicine to human-computer interaction, MIT students built a hackathon prototype that’s equal parts fascinating and unsettling: a wearable that can gently “steer” your hand using AI plus electrical muscle stimulation. A head-mounted camera feeds what you’re looking at to a vision-language model, the system interprets what you ask for, and then it triggers small electrical pulses on your arm to activate specific muscles—nudging your wrist or fingers into motion. In demos, it guided simple gestures and even basic piano notes. As a 48-hour student build, it’s very much a proof of concept, not a polished product. But the broader idea is worth watching: AI that doesn’t just advise you on a screen, but can help you learn physical actions, support rehabilitation, or assist people who need help translating intent into movement.

Possible US tariffs on chips

Finally, a policy note with major industry implications: the Trump administration is again floating the idea of tariffs on imported semiconductors as a lever to encourage more chip manufacturing in the United States. The U.S. Trade Representative, Jamieson Greer, emphasized that nothing is imminent and that discussions with the industry are ongoing. Even so, the signal matters. Chips sit at the center of everything from cars to data centers, and tariffs can quickly ripple into pricing, supply stability, and investment decisions. For businesses, the key word here is uncertainty: markets don’t need new tariffs to feel the impact—sometimes the possibility alone changes planning.

That’s the tech landscape for May-23rd-2026: AI nudging at the boundaries of mathematical discovery, reshaping how we search, and turning up in everything from smart homes to protein design—while policymakers consider moves that could rewire chip supply chains. If you want, I can also bundle today’s stories into a quick shareable briefing for your team or class. Thanks for listening to The Automated Daily, tech news edition. I’m TrendTeller—see you next time.

AI breaks Erdős geometry conjecture & Google search goes more AI - Tech News (May 23, 2026)

Our Sponsors

Today's Tech News Topics