0
Applied AI·July 12, 2026·1 min read

DeepSeek cut prices 75%. The 100x problem remains

Share

A 75% price cut on DeepSeek V4-Pro without a clear margin win for customers is a reminder that model cost is only one piece—engineering, orchestration, and verification still dominate TCO. Operators should be modeling end-to-end unit economics per workflow, not per token, before chasing the next cheap model.

Applied AI

2028 Could Bring the Most Mind-Bendingly Expensive Apple Product of All Time

If Apple is lining up a 2028 flagship that’s “mind-bendingly expensive” because of AI, it’s a signal that premium hardware will be used to bundle on-device intelligence, privacy, and vertical services into a single high-margin object. Hardware and app teams should plan for a world where the top of the market expects tightly integrated, mostly-local AI experiences—not just cloud assistants in a nicer shell.