Vercel Acquires Better Auth to Give Agents an Identity

Vercel Acquires Better Auth to Give Agents an Identity

Vercel has acquired Better Auth, the open-source TypeScript auth library pulling 4.7M+ weekly npm downloads and built by 850+ contributors [1]. The full team is joining Vercel, with an explicit mandate: build identity infrastructure not just for human users, but for AI agents acting on their behalf [2].

This is a bigger deal than it looks. As agents increasingly take actions — booking, purchasing, deploying code — "who is this agent, and on whose authority" becomes a hard infrastructure problem, not a philosophical one. Developer reaction on X and Hacker News was largely positive, with builders framing it as Vercel filling a gap nobody else was moving fast enough on [3].

Auth was already unglamorous, necessary plumbing. Agent identity is about to be the same — except get it wrong and you've got an autonomous system with the wrong permissions acting at machine speed. Expect more acquisitions in this lane before the year is out.

Google's AI Scientist Automates the Research Pipeline

At ICML 2026 in Vienna, Google Research presented a multi-agent system — AI Scientist, alongside a "ScientistOne" variant — that automates hypothesis generation, experiment design, and paper synthesis end-to-end [1][2]. Supporting tools include a ScholarPeer reviewer agent and a PaperBanana visual generator, essentially closing the loop from idea to publishable output [3].

The research community's reaction on X was a mix of genuine excitement and unease — demos were widely shared and praised, but the subtext is unavoidable: if agents can generate, review, and illustrate research autonomously, the human role shifts hard toward judgment — deciding what's worth investigating, and whether the output is trustworthy — rather than execution.

That's the pattern showing up everywhere this week, from coding to science: the work of doing is being automated: the work of deciding isn't.

Europe Pushes for AI Sovereignty Amid US-China Tensions

Tensions between Anthropic and Alibaba over alleged capability extraction, plus tightening U.S. export controls, have reignited Europe's push for AI sovereignty [1][2]. Bruegel and others are calling for allies to actively plan around dependency risk rather than assume continued open access to U.S. or Chinese frontier models [3].

Nordic and EU voices on X are increasingly vocal about building local alternatives — not out of protectionism alone, but because geopolitical access risk is now a real operational concern for any company betting its stack on a foreign model provider. For Nordic AI builders specifically, this is the strongest signal yet that "just use the best API" is not a strategy — it's a dependency you need a plan B for.