AI cracks Erdős geometry puzzle & Google blends AI into search - Tech News (May 24, 2026)

Let’s start with that headline-grabbing claim from OpenAI: it says one of its experimental reasoning systems found a better answer to a classic geometry question tied to Paul Erdős. The “unit-distance problem” is basically about arranging points on a plane to maximize how many pairs are exactly one unit apart. Erdős proposed a construction back in 1946 and dared the field to beat it. OpenAI now says its AI produced an arrangement that outperforms Erdős’s long-standing conjecture—and importantly, independent mathematicians not affiliated with the company reportedly checked and verified the result. What makes this interesting isn’t just the bragging rights. It’s the possibility that AI systems can stumble into genuinely new math—using advanced tools, long reasoning chains, and a style of exploration that looks less like autocomplete and more like creative search. OpenAI hasn’t named the model, and the full write-up isn’t public yet, so the community will still want details. But if this holds up under broader scrutiny, it’s a meaningful marker: AI not just explaining known results, but helping extend the frontier.

Staying in AI—but moving from proofs to daily life—Google is redesigning its iconic search box to better fit an AI-first way of asking questions. The box can stretch to handle longer, more conversational prompts, and it’s being tuned for multimodal input—meaning you can mix in things like images or other files to guide what you’re searching for. The bigger shift is how Google continues blending AI summaries with traditional blue links. The company’s bet is that people want a quick synthesized answer plus paths to dig deeper. Critics, meanwhile, worry this pushes the web toward a more closed experience: fewer clicks out to publishers, fuzzier accountability for where answers come from, and higher stakes when an AI summary is wrong or overly confident. In other words, it’s not just a UI tweak—it’s a power shift in how information gets packaged and paid for online.

Now for a privacy story that deserves real attention: researchers at Germany’s Karlsruhe Institute of Technology say ordinary WiFi networks can be used to identify people with near-perfect accuracy by analyzing how radio waves reflect off the human body. The unsettling part is the data source: they rely on routine beamforming feedback that devices send to routers during normal operation—and that information is often unencrypted and readable by someone nearby. In tests with a large group of participants, they report identification that works in seconds, even if the person isn’t carrying a device, because nearby connected gadgets still generate enough WiFi activity. If that generalizes, it turns common routers in homes, offices, and public venues into potential silent tracking infrastructure. The researchers are pushing for privacy protections to be baked into upcoming WiFi standards, because this isn’t about one vulnerable product—it’s about a capability that could scale quietly and fast.

On the business and policy side, quantum computing is getting another push from Washington. The U.S. Department of Commerce has signed a set of letters of intent tied to CHIPS and Science Act incentives, spreading support across multiple quantum companies and approaches. The takeaway isn’t that quantum is suddenly “solved.” It’s that the U.S. is trying to de-risk the race by funding a portfolio rather than betting on one hardware strategy—and markets noticed, with quantum-related stocks reacting quickly. For the broader tech world, this matters because it’s a signal: quantum is moving from a mostly research-driven storyline to a commercialization storyline, even if major hurdles like reliability and scale are still very real.

That funding news also intersected with a broader market moment: Nvidia just capped off the latest tech earnings cycle with another huge quarter and a fresh wave of shareholder-friendly moves. But the more revealing part was strategic. CEO Jensen Huang acknowledged Nvidia has largely ceded China’s AI chip market to Huawei under the weight of export controls and China’s domestic push. That’s a blunt reminder that geopolitics is no longer a side plot in the semiconductor story—it’s shaping growth paths. Nvidia also continues signaling it wants to be more than the company powering hyperscale data centers. By changing how it reports parts of the business and emphasizing “edge” opportunities—think PCs, robotics, and cars—it’s telling investors the next phase is about where AI runs, not just where it’s trained. And as expectations climb, even blowout numbers don’t guarantee the stock reaction you might assume.

Finally, two advances in lab-grown human tissues show how quickly organoid research is maturing—literally. First, researchers described a “confined culture” approach that helps fuse large numbers of stem-cell-derived gut spheroids into longer, tube-like tissues that can be transplanted earlier and engraft more effectively than typical small, round organoids. The headline result: these engineered gut tissues developed their own human-origin enteric nervous system—neurons and supporting cells—without needing extra added nerve precursor cells. After transplantation, they even showed nerve-dependent contractions that resemble adult intestinal movement. That’s a big step for disease modeling, and a promising direction for future graft-based treatments for intestinal failure. Second, a team in Shanghai reported what they describe as the first lab-grown sinoatrial node organoid—the tiny natural pacemaker region of the heart. Because this tissue is hard to study directly in patients and animal models don’t always match human pacing, a human-like organoid could be valuable for understanding arrhythmias and testing drugs. Longer-term, it points toward the idea of biological pacemakers that might reduce dependence on implanted electronics—still a future concept, but one that’s becoming easier to imagine.