Former Intel CEO on What Went Wrong, What's Next + Lovable CEO on the Real Promise of Vibe Coding

The first half is a postmortem on Intel from the man who joined at 18 and later ran it. Gelsinger's core diagnosis: a company founded by deeply technical leaders — Grove, Moore, Noyce, an executive staff where 15 of 20 were PhDs — spent roughly 15 years being run by business people, who promote more business people. In the five or six years before he returned as CEO, Intel handed $100 billion to shareholders in buybacks and dividends while going a decade without building a new factory or buying EUV machines. He walks through the three misses from the inside: Steve Jobs quietly porting Apple's operating system to x86 four releases before anyone at Intel knew silicon independence was coming; Nvidia's CUDA improving release by release while Intel scoffed at graphics cards and killed its Larrabee competitor a week after Gelsinger's first departure; and TSMC turning the foundry model Intel didn't care about into five times Intel's wafer volume — now more like seven.

On geopolitics he is blunt: Taiwan holds less than three weeks of energy reserves, so a blockade — which China has rehearsed repeatedly in recent years — could brown out the island without a shot fired, and a fab that loses power stays down for 90 days. The economic impact, he says, would exceed the Great Depression. The CHIPS Act has moved US leading-edge share from about 12% to 18%. On the AI bubble question, he argues energy capacity is the built-in ceiling that keeps the buildout from running too far ahead of itself, predicts a two-decade buildout rather than a two-year one, wants cost per token down by five orders of magnitude, and forecasts meaningful quantum results across multiple industries before 2030.

The back half is Jason Calacanis interviewing Anton Osika of Lovable, 20 months into the company: a million new projects built weekly, more than 50 million apps on the platform, 700 million monthly visits to hosted applications, and 500 million in revenue reached in May. Four out of five users are non-technical, and Osika describes customers running businesses past a million dollars in revenue on the platform — including a Nursa employee who built a nurse-certification product and then replaced more than ten internal tools, saving over a million dollars a year. Calacanis tells his own story of an employee building the company intranet in hours without permission. Osika sketches the roadmap — hosting as a second product line growing faster than building, and an 'AI co-founder' that analyzes your business overnight — and explains the model strategy: route each request to the best frontier or open-weight model, with a Stockholm post-training team fixing the specific mistakes frontier models make.