OpenAI and Anthropic alumni are spawning a new wave of AI startups in Silicon Valley
Silicon Valley is reviving an old term for a new group: the "mafia" effect once associated with PayPal is now being applied to former employees of OpenAI and Anthropic. TechFlowPost argues that, through the first half of 2026, enough people have left the two leading AI labs to build a long roster of founders launching new companies. Unlike the PayPal era, most are not trying to build frontier models from scratch. Instead, they are clustering around the areas large labs have left open: automated AI research, systems that help models improve themselves, verifiable math, agent workflows for enterprises, personal AI, AI-native hardware, and safety evaluation. The report names several examples. Former OpenAI research VP Jerry Tworek launched Core Automation to build an automated research lab. Former Anthropic researcher Behnam Neyshabur and others started Mirendil, which reportedly raised $200 million for self-accelerating systems that let models help improve models. Former OpenAI researcher Jesse Han founded Math Inc, focusing on machine-verifiable mathematical proofs. Other startups highlighted include Rational, Zavify, River AI, Blackstar, Syntony, Resolution, and Guidelight. TechFlowPost says the timing is not accidental. OpenAI arranged a secondary share sale last autumn that let employees cash out $6.6 billion at a $500 billion valuation, while both OpenAI and Anthropic are said to be preparing for IPOs. Venture capital is already following the talent. Still, the article warns that dozens of startups are crowding into the same narrow opportunities, and most may not survive.
