What building Tesla's autopilot chips taught Cognition's President about AI
Transformer architecture standardization will enable extreme chip specialization and massive inference performance gains
“I think we're gonna have crazy levels of specialization because there's gonna be so much inference demand. And that's gonna unlock just like super exciting levels of performance.”
Cognition's President recounts how Tesla stripped division support from autopilot inference chips to maximize thermal headroom and throughput, since convolutional layers never needed it at inference time. The lesson applied forward: as transformer architectures stabilize, chip designers can make increasingly aggressive assumptions and remove unused operations. The speaker argues this will drive a wave of specialized inference hardware with outsized performance improvements.