Subscribe
Share
Share
Embed
Daily AI news and research, distilled. UpNext AI breaks down the most important developments in artificial intelligence—from major industry moves to cutting-edge papers.
Welcome to the UpNext AI podcast. It's Wednesday, May 20th, 2026, and here's what matters in AI today.\n\nFirst up, Google. According to the Financial Times, Google plans to release smart glasses and add AI agents to its search engine, with the new features powered by a Gemini model. The broader point here is competitive positioning. FT reports that Sundar Pichai framed the move as part of Google’s effort to close the gap with Anthropic and OpenAI.\n\nWhy this matters is that Google is not treating AI as a side feature anymore. It’s weaving Gemini into core consumer surfaces: search, where Google already has massive reach, and wearables, where smart glasses have long felt promising but unfinished. If Google can make AI agents useful inside search, that changes the product from a place you visit for links into something closer to an active assistant. And if smart glasses return with better AI, Google gets another shot at ambient computing, this time with far stronger models behind it.\n\nWhat we do not have yet from the reporting are the detailed technical capabilities, product names beyond the category, or a deeper rollout schedule. But even at this level, the message is clear: Google wants AI to show up not just in chat windows, but across search, devices, and everyday tasks.\n\nOur second story is OpenAI, which announced OpenAI for Singapore. The company says it’s a multi-year partnership aimed at expanding AI deployment, building local talent, and supporting businesses and public services with AI.\n\nThis is narrower than the Google story, but it’s still important because it shows how the competition is spreading beyond models and apps into country-level adoption programs. OpenAI is not just selling access to tools here. The company is describing a longer-term effort around workforce development, deployment, and public sector use.\n\nThat makes Singapore a useful signal. When AI companies talk about national partnerships, they’re trying to become part of the local infrastructure for education, business, and government services, rather than just another software vendor. In this case, OpenAI’s announcement stays high level. It does not spell out funding, named partners, or implementation details in the summary we have. Still, the direction is notable: frontier AI firms are increasingly competing to become embedded at the ecosystem level, not only at the product level.\n\nNow to the research story. A new Nature paper, titled A multi-agent system for automating scientific discovery, describes a system called Robin. The paper says scientific discovery is an iterative cycle of observation, hypothesis generation, experimentation, and data analysis. Robin is designed to help automate that loop, rather than only assisting with one piece of it.\n\nThe paper goes further than a generic workflow demo. According to the article text, Robin integrates literature search agents with data analysis agents so it can generate hypotheses, propose experiments, interpret results, and then generate updated hypotheses. The researchers say they applied it to experimental biology and used it to identify potential therapeutic candidates for dry age-related macular degeneration.\n\nThe specific examples in the paper are striking. Robin proposed enhancing retinal pigment epithelium phagocytosis as a therapeutic strategy, identified ripasudil and KL001 as candidates with confirmed in vitro efficacy, and then proposed and analyzed a follow-up RNA-seq experiment. The paper says that follow-up work pointed to upregulation of ABCA1 as a possible novel target. The researchers also say ripasudil is a clinically used ROCK inhibitor that had not previously been proposed for treating that condition.\n\nThe big takeaway is not that AI has replaced scientists. It’s that researchers are starting to test AI as a coordinated research teammate that can keep the discovery loop moving from paper search to hypothesis to experiment planning to analysis. We do not have broader performance data here, and the entry does not include external validation beyond the paper itself. But if systems like this hold up outside a single study, the practical implication is huge: AI could help labs speed up the slow, repetitive parts of turning data into experiments and experiments into new ideas.\n\n...Are you building apps with voice? Elevate your app's voice capabilities with ElevenLabs. Their API is a game changer for embedding dynamic, responsive voice interactions in your applications, providing unprecedented realism, flexibility and latency. In fact, you're listening to one of their voices - right - now. If you are a developer looking to elevate user experience with natural voice interfaces, this is your solution. Visit up next dot fm slash eleven to check out their latest offerings. ...\n\nIn headlines, OpenAI also announced the next phase of its Education for Countries program. The company says it is expanding AI adoption in schools through new partnerships, teacher training, and tools intended to improve learning outcomes. That fits with the broader pattern we just discussed: AI firms are trying to become part of national infrastructure, in this case through education.\n\nNext, TechCrunch reports that Google is transforming Search from a list of links into an AI-powered experience with conversational answers, autonomous agents, and more interactive interfaces. One likely consequence, according to the report, is even more pressure on publisher traffic across the web. So the search shift is not just a product story. It is also a business model story for the internet itself.\n\nAnd one more in science. India Today reports that SandboxAQ, chaired by former Google CEO Eric Schmidt, has partnered with Anthropic to bring SandboxAQ’s scientific models into Claude. The report says the goal is to give researchers access to drug discovery and materials science tools through a conversational interface, rather than requiring specialized infrastructure. India Today also reports that SandboxAQ has raised more than 950 million dollars. The larger idea here is that frontier chat models are increasingly being paired with domain-specific scientific systems, especially in drug discovery.\n\nBefore we wrap up, a quick note: this podcast is generated with the assistance of AI and is intended for informational purposes only. All referenced articles, research, and commentary remain the property of their original authors and publishers.\n\nIf you enjoyed this episode, don't forget to subscribe, rate, and leave us a review! And that's your briefing for today. Full source links are in the episode notes, and we'll be back tomorrow with what's up next!