Was looking through old chats from 10 years ago about Cerebras’s Series B .. such a full circle. in 2017, if you bet on GenAI you were a "vibe investor" - Academia hated on Transformers for not being "elegant" enough for CV (RIP ViT/DETR haters). The NLP scientists called decoder-only models a dead end. 2017 was one of those quiet historical forks, no one in the industry grasped how deeply it would shape what became the Fourth Industrial Revolution Now, the same people are begging for H100s. Meanwhile, Cerebras is still fighting the physics of tiny SRAM and no HBM. Hardware is hard; being right early is harder. and ofc Cerebras didn’t feel obvious to me at all Looking back, my judgment was full of the constraints of that moment: I wasn’t sure Nvidia could become the backbone of hyperscale AI datacenters. Google still looked like the kind of company that would keep its best infrastructure for itself. And Cerebras’s wafer-scale engine felt, to me, like a very expensive piece of hype. That’s the thing about living through history in real time..the future rarely sounds loud when it first arrives. 2017 to 2026 was AI’s version of Hemingway: slowly, then suddenly all at once. And four years earlier, 2014 to 2023 was crypto’s The first 7 years of every exponential industry are usually just a long muted prologue, then last 3 tear through the world as a full symphony That’s still the most intoxicating thing about spending the last 15 years around frontier tech: getting to touch history in draft form, watching founders stay faithful while lost inside it, and watching random choices separate people from fortunes they couldn’t even imagine yet Slowly but surely, suddenly all at once