For a while I kept asking whether to stay in AI. By the third time I noticed it was a misplaced question — disguising an internal judgment (where on the value chain to place myself) as an external one (whether the industry is worth being in). This essay isn't about whether to stay. It's about the real question hidden underneath — once model capabilities keep climbing and engineering and interface keep getting commoditized, what is left in the product layer that won't be flattened.
An observation that started with a floating-lyrics music player. As scheduling, scaffolding, engineering, and interfaces all collapse into AI, the three lines software has always run on — processing, providing, and producing data — are heading for very different fates.
In 1960, Licklider wrote Man-Computer Symbiosis and imagined two living systems locked in mutual dependency. Sixty-six years on, the door has still not opened — we have ever-better tools, but never a partner.