Blog
Bezos Bets $38B on AI While Product Managers Wonder If They Still Have a Job
Jeff Bezos' Project Prometheus raises $10B at a $38B valuation, Anthropic ships Claude Design for non-designers, and product managers admit AI is making their role both easier and more existential.
Published April 21, 2026
Bezos' secret AI project is now worth more than most unicorns combined
Jeff Bezos' Project Prometheus is raising $10 billion at a post-money valuation of around $38 billion. That's not a typo. The secretive AI startup he cofounded in November is pulling in mega-round money barely five months after launch, and we still don't know what it actually does.
The round is still in progress, so the final number could shift, but the trajectory is clear: if you're Bezos and you want to build an AI company, capital is not the constraint. The details remain sparse—nobody's talking about the product, the team size, or even the specific problem being solved. What we do know is that this valuation puts Prometheus ahead of most venture-backed companies that have shipped actual products people use.
It's the kind of financing that makes seed-stage founders checking their bank balance feel a certain way.
Anthropic wants to help you make slides without a designer
Anthropic launched Claude Design, a new product aimed at people who need visuals but don't have a design background. Think founders building pitch decks, product managers mocking up features, or anyone who's ever said "I can picture it, I just can't draw it."
The pitch is straightforward: describe what you want, and Claude generates a quick visual. It's not replacing Figma or a contract designer for production work, but it's aimed squarely at the "good enough for the next meeting" tier. That's a real use case—plenty of early-stage teams spend more time wrestling with layout tools than they'd like.
If it works as advertised, it's another step in the "AI as co-pilot for non-specialists" direction. Whether it actually saves time or just creates a new category of slightly-off visuals that still need manual cleanup is the open question.
Google keeps expanding Gemini's footprint
Google is rolling out Gemini in Chrome across seven new countries: Australia, Indonesia, Japan, the Philippines, Singapore, South Korea, and Vietnam. The feature lets users interact with Gemini directly from the browser, continuing Google's strategy of embedding its models everywhere a text box exists.
Meanwhile, Google AI Studio now offers expanded usage limits for AI Pro and Ultra subscribers. If you're paying for access, you get more tokens and fewer rate-limit headaches. It's an incremental update, but it signals Google is trying to make the developer experience less punishing for people who actually want to build with these tools.
A wave of funding for things that might matter
Collov Labs raised $23 million in Series A for visual AI that turns images and camera inputs into actions. The company is betting that computer vision plus LLMs can automate workflows in industries where "look at this and do something" is the core task. Think quality control, logistics, field service—anywhere a human currently eyeballs something and makes a call.
On the consumer side, Bond launched out of stealth with $5 million in seed funding. The app uses AI to help users catalog memories and spark plans with close friends. It's built by ex-VC Dino Becirovic and former Meta engineer Hanxin Jin, and the pitch is that AI can surface the right memory or suggest the right hangout at the right time. Whether people want an algorithm nudging their social calendar is an experiment in progress.
Product managers are having a moment (and not the good kind)
A new report says the AI boom is both energizing and exhausting product managers. On one hand, tools like Claude let PMs prototype and test ideas without waiting on eng resources. On the other hand, the coordination overhead hasn't gone away—it's just shifted. Now you're managing AI tools and the fallout when they don't quite work as expected.
The role is more engaging because building is faster. It's also more stressful because the pace is unrelenting and the "just ship it" pressure is higher when everyone assumes you can iterate in hours, not sprints. It's the kind of dynamic that makes the job feel both more important and more replaceable at the same time.
The week in brief
Yelp's updated AI assistant can now answer questions about a restaurant and book a table or place an order in one conversation. It's the kind of feature that feels obvious once it exists—why should "find a place" and "reserve a table" be separate tasks?
Elsewhere, Reverie announced conversation forking and multi-character chats for its NSFW character AI platform. If you're building in the AI roleplay space, you're now competing on features like branching narratives and group interactions. It's a niche, but it's a niche that apparently has enough users to justify a feature roadmap.
The Monday vibe this week: big money chasing unclear bets, incremental updates to tools people already use, and a growing sense that we're all just trying to keep up.