Follow your tools, not a fixed dream
A direction can stay serious without pretending that the world—or the person choosing it—will stay still.
We are often taught to choose a distant identity early, then treat persistence as the proof that the choice was correct. But knowledge changes through practice. The world changes even faster. A goal chosen with incomplete information should be allowed to evolve when our information improves.
For me, the more useful question is not “What single thing must I become?” It is “What can I make with the tools, values, curiosities, and opportunities I have now?”
The toolkit is not only technical
Python and diffusion models are tools. So are drawing, aesthetic judgment, curiosity about philosophy, the patience learned from research, and the instinct to make something interactive instead of leaving it as an idea. A toolkit is partly what you know and partly what repeatedly pulls your attention.
Project-based learning works because every real project exposes a missing tool. Even a failed attempt can enlarge the set of things the next attempt can become.
Direction without rigidity
This is not an argument for drifting without standards. The problem still has to matter. The attempt still has to be honest. But the path can behave more like an agile process than a twenty-year waterfall plan: make, review, learn, and change direction when reality teaches you something.
Understand the problem space. Understand the toolkit. Then follow the slope where they meet.