microsoft didn't ban engineers from using AI. they moved them from claude code to their own tool to control costs. that's vendor consolidation, not AI rejection. uber's 84% adoption with 70% of committed code from AI isn't a failure story - it's a pricing problem. the value was real. the unit economics weren't. these are two very different things. token prices dropped 75% in the last 6 months alone. deepseek just made their permanent pricing 4x cheaper. gemini 3.5 flash is frontier quality at flash prices. the cost curve is collapsing faster than the adoption curve is rising. this thread takes a snapshot of today's pricing and projects it forward as if nothing changes. everything about AI pricing is changing every quarter.

Personal update: I've joined Anthropic. I think the next few years at the frontier of LLMs will be especially formative. I am very excited to join the team here and get back to R&D. I remain deeply passionate about education and plan to resume my work on it in time.