Some people talk about the future of AI like it’s a fairy tale. They open their mouths and it’s always the same lines: “Tokens will get cheaper,”
“Compute will drop exponentially,” as if startups just need to lie on the beach and wait for 10×, 100×, 1000× cost reductions to fall from the sky. Yes, the *trend* is real. But the *timeline* is a black box.
Costs *will* go down—maybe in two years, maybe in eight, maybe right after your company is already dead. Most startups simply won’t live long enough to see that “technical inflection point.”
Either you raise a huge pile of money and burn your mistakes into tuition, or you face reality and make money early with limited resources. Optimizing token usage isn’t embarrassing—it’s basic survival. People who mock it just don’t understand how brutal the real world is.
Then there’s another group who keeps shouting that “chat interfaces will die.” As if hundreds of thousands of years of human linguistic instinct would vanish just because AI iterates fast. Language is the lowest-friction, highest-bandwidth interface humans have.
It’s not a UI layer—it’s built into our biology. The product form will absolutely evolve: from big chat boxes to voice-first interactions, contextual agents, implicit intent, and background automation.
But that’s not “chat disappearing.” It’s “conversation being absorbed into everything.” It becomes everywhere and invisible, yet far more powerful. Saying chat will vanish is like saying “mobile computing will kill keyboards.” Keyboards changed, sure—but they never disappeared.