Anthropic researcher expects the world to end in six months, Delphi host says

Why it matters

Arrington's view puts a longtime early-stage investor's core question - what advantage can a startup keep? - against AI's fast-moving model competition, while making clear that his skepticism is a thesis rather than proof.

Tommy Shaughnessy, host of The Delphi Podcast, said an unnamed Anthropic researcher told him he was "enjoying my last six months" during dinner at Shaughnessy's home. In an episode released October 9th, Shaughnessy said the researcher was serious and that "most people in Anthropic" believed they had six months. He added that the researcher was the only person from one of the leading AI labs he had spoken with in person. The account is a secondhand anecdote; the episode provides no independent evidence for the claim.

Michael Arrington, the founder of TechCrunch and an early investor in Uber and Airbnb, asked whether the researcher was joking. Shaughnessy said he was not. The conversation then turned to whether Anthropic could stop developing AI: Shaughnessy argued that other companies and China would continue regardless. Those remarks reflect the speakers' views, not evidence that catastrophe is imminent or that Anthropic employees broadly share the reported belief.

The exchange came amid a wider discussion of the AI race. Shaughnessy argued that a US lab slowing development would risk ceding ground to China, while Arrington described the situation as a "catch-22" in which companies keep advancing despite uncertainty about what happens next. Arrington also characterized calls for leading labs to set the pace of regulation as potential regulatory capture, arguing that restrictions could protect incumbents from competition.

Arrington's own investment thesis was more focused on competition: "Moat building in AI seems to be a difficult thing to do," he said. The interview offers no market data to test that view. Arrington's point was that the usual startup playbook of building a technical lead may be harder to rely on when competitors can release capable models openly and the leading systems keep changing.

The episode also ranged across founder psychology, crypto and Arrington's history in tech. He recalled investing "like a hundred grand" in Airbnb before he was sure it would succeed, saying the investment ultimately made substantial returns for his limited partners. The anecdote is self-reported and does not establish the fund's return multiple.

Arrington founded TechCrunch in 2005 after working as a securities lawyer, according to his TechCrunch biography, which lists Uber, Airbnb and Pinterest among his venture investments. The episode arrived on the 20th anniversary of Arrington's report that Google had agreed to buy YouTube for $1.65 billion in stock. His October 9th, 2006 TechCrunch story reported the deal as it was confirmed.

Shaughnessy asked how Arrington evaluates founders and builds teams around his own weaknesses. Arrington said he values hiring well and spotting early-stage companies likely to win, while acknowledging that managing people is a weakness he has addressed by hiring operators. He also said he considers AI more revolutionary than crypto, even as he continues to argue for Bitcoin as hard money. The Delphi Podcast episode runs 1 hour and 16 minutes and covers China, founder psychology, the sale of TechCrunch, market cycles and Bitcoin.