
Computer manufacturing is--after cars, energy production and illegal drugs--the largest industry in the world, and it's one of the last great success stories in American business. Accidental Empires is the trenchant, vastly readable history of that industry, focusing as much on the astoundingly odd personalities at its core--Steve Jobs, Bill Gates, Mitch Kapor, etc. and the hacker culture they spawned as it does on the remarkable technology they created. Cringely reveals the manias and foibles of these men (they are always men) with deadpan hilarity and cogently demonstrates how their neuroses have shaped the computer business. But Cringely gives us much more than high-tech voyeurism and insider gossip. From the birth of the transistor to the mid-life crisis of the computer industry, he spins a sweeping, uniquely American saga of creativity and ego that is at once uproarious, shocking and inspiring.
> Look at his last few blog posts, outlining a radical alternative to LLMs. Read closely, and at the heart of that idea we find "assume enterprise search".
I found the patent for his company's technology (https://patents.google.com/patent/US20250259069A1). I always have trouble reading patents, but as far as I can tell it's basically a wrapper around commodity LLMs. The reliability improvements come from prompting the LLM to only source its data from an external database that you provide rather than from whatever it was trained on, and from prompting multiple LLMs simultaneously then doing semantic analysis on the responses to make sure they all agree. I imagine that doing this would probably reduce hallucinations, but there's no hard constraint to prevent them entirely and I'm not sure what the moat is.
