Latest / UpNext AI / H1’s Healthcare Data Bet, Anthropic’s $65bn Raise, and Self-Evolving Agent Skills | UpNext AI – May 29, 2026
Transcript
- Welcome to the UpNext AI podcast. It's Friday, May 29th, 2026, and here's what matters in AI today. First up, TechCrunch reports that healthcare data platform H1 has secured $40 million from CVS Health Ventures. The bigger point here is the kind of company still getting funded. H1 co-founder and CEO Ariel Katz argues that AI may be able to replicate workflow SaaS, but not a business built around unique underlying data. In H1’s case, that means detailed doctor data sold into pharma, hospital systems, and health insurers. TechCrunch says H1 was not actively looking to raise, had turned cash flow and EBITDA profitable last year, and expects to grow more than 40 percent this year. The story also notes H1 was last valued at $750 million when it raised $100 million in 2021. So the signal is not just fresh funding. It’s that investors still seem willing to back companies that can make a credible case for a durable data moat, especially in healthcare. Next, the Financial Times reports that Anthropic has finalized a $65 billion funding deal, valuing the company at $965 billion including the new money. Even without the full article text in our packet, the scale makes this notable. It suggests markets are still willing to fund frontier model labs as strategic infrastructure plays, not just software companies. And in the FT’s framing, the round would put Anthropic above OpenAI by valuation. At minimum, it is another sign that the race for top-tier AI models is still pulling in extraordinary amounts of capital. For the research note today, from earlier this week on arXiv, a paper called MUSE-Autoskill looks at how agents might improve by building reusable skills over time. The paper argues that current LLM agents often treat skills as isolated and static, which limits reuse, reliability, and long-term improvement. Its proposed framework pairs skill creation with memory, management, and evaluation, so an agent can create, store, reuse, and refine skills under one lifecycle. In plain English, the idea is that agents should not just remember facts. They should remember how to do things, and get better at those methods over time. We do not have benchmark results in the packet, so the main value here is the direction of travel. Bottom line: if agents are going to handle longer, messier workflows, they will likely need durable skill memory, not just better prompts. ...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. ... TechCrunch reports Anthropic has released Opus 4.8 with a new tool called Dynamic Workflows, aimed at coordinating swarms of subagents. OpenAI has published a Frontier Governance Framework, which the company says aligns its AI safety, security, and risk practices with emerging EU and California regulations. Developer note: Simon Willison released llm-anthropic 0.25.1 with support for Claude Opus 4.8, plus a fast-mode option for eligible organizations and model-specific max token defaults. And The New York Times’ The Daily published an episode yesterday titled Can A.I. Make People Feel Less Lonely? Host Rachael Abrams and colleague Eli Saslow look at how AI is being used to help older Americans combat loneliness and maintain independence, including a robot companion called ElliQ, while also exploring the concerns that come with that kind of companionship technology. Before 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. If 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 Monday with what's up next!