News of the day

1. Anthropic's IPO prospectus details existential AI risks, especially recursive self-improvement, while the company races to close that loop for competitive advantage. → Read more

2. Reddit's Blair Metcalf details its internal AI prototyping platform, and Claire tests Anthropic's Claude Duet personal agent in this episode. → Read more

3. Weekly AI roundup covering SpaceX's $40B GPU debt deal, SoftBank's $50B Malaysia plan, partisan bias findings and upcoming big tech earnings. → Read more

4. Slush 2026 in Helsinki brings AI agents, EU leaders and top investors together to shape where European startup funding goes next. → Read more

Our take

Hi Dotikers!

There is something almost poetic about a company telling Wall Street "please give us billions" and "this technology might end badly for everyone" in the same document. That is exactly what Anthropic just did. Its IPO prospectus dedicates a staggering 94 pages to risk factors, far more space than it gives to explaining how it actually makes money. And these are not your usual boilerplate warnings about currency fluctuations or supply chains. Anthropic explicitly flags recursive self-improvement, the scenario where AI systems take over AI research itself and start advancing faster than human evaluators can follow.

Here is the twist that makes this filing genuinely fascinating: the very thing Anthropic warns about is also the thing it sells as its competitive edge. The company calls it "closing the loop" and presents it as a structural advantage in the race for frontier AI. So the pitch to investors is essentially: this could spiral beyond our control, and that is precisely why you should fund us. It takes a certain audacity to put the apocalypse in the brochure and call it a growth strategy.

Let's be honest, this is less a confession than a brilliant legal shield. By disclosing everything, Anthropic protects itself from future lawsuits while positioning itself as the responsible adult in the room. Smart lawyering, yes. But it also exposes the core contradiction of the entire industry: everyone agrees the risks are real, several bills in Congress, including one backed by Bernie Sanders, propose banning recursive self-improvement, yet nobody wants to slow down first. The financial stakes are simply too enormous for restraint to win on its own.

The uncomfortable truth is that we are now relying on companies to regulate themselves while they sprint toward the very thing they fear. A prospectus is not a safety plan. At some point, lawmakers will have to decide whether warnings printed in an SEC filing count as doing something.

G.

Leave Granola and get up to 12 months free of Wispr Flow Notetaker + Dictation

If you have paid time left on an individual Granola plan, we'll match it with a Wispr Flow subscription that includes Notetaker and dictation, and add bonus time, up to 12 months total. Sign in or create a Wispr account and submit proof of your plan to check eligibility.

Meme of the day