Why small beats big, every single time.
Most creators follow this loop:
Big idea → Excitement → Planning → More planning → Overwhelm → Abandon
Then they feel bad about themselves. Then they get a new big idea. Repeat.
The fix isn't more discipline. It's smaller projects.
They get finished. A finished tiny project beats an abandoned big one. Every time.
They build momentum. Finishing one small thing makes it easier to start the next one.
They compound. 10 tiny projects = a portfolio, a body of work, proof that you ship.
They teach you more. You learn more from shipping 5 small things than planning 1 big thing.
-
Define "done" before you start. If you don't know what finished looks like, you'll never get there.
-
Scope it to one sitting. If it can't be finished in 1-2 hours, it's too big. Cut it in half.
-
No scope creep. When you think of a cool add-on, write it down as a future seed. Don't add it now.
-
Ship the ugly version. Version 1 just needs to exist. It doesn't need to be good.
For ML devs specifically:
- A tiny project = one notebook, one model, one insight
- A tiny post = one experiment result + one sentence of reflection
- A tiny contribution = one PR, one bug fix, one doc improvement
You don't need a research paper. You need a shipped thing.
You're not bad at finishing things. You've just been picking projects that are too big.
Start smaller than feels reasonable. Finish it. Feel good. Start the next one.
That's the whole philosophy.