I've been thinking about my system of execution. Not in some grand productivity-framework way, but more like: why do I keep getting stuck?

Here's the thing. My thought process is fast. I can switch contexts quickly, grasp new ideas, connect dots across domains. That's never been the problem. The problem is that I don't move. I get a spark - say, "let's work on few-shot classification" - and instead of working on it, my brain teleports. Where did this idea come from? How did it evolve? Who worked on it first? What were the original constraints? How did they avoid this particular failure mode? I trace the entire genealogy of the idea before I've written a single line of code.

It's a rabbit hole disguised as learning. And it's eating my time.

The Biryani Principle

Here's what's funny. When I crave biryani, I don't do any of this.

I don't sit there wondering about the Mughal origins of the dish. I don't trace back to which regional variation came first. I just think: it's been a while, let's make biryani. Then: when do I have time? Tomorrow morning. Then I spend maybe ten minutes planning - what do we need, what do we have, what's missing? If something's missing, I walk to the store or order it. And then I cook.

Trigger. Quick plan. Handle the gaps. Execute. Done.

That cycle - from craving to cooked biryani - is fast, decisive, and honestly? It produces great results. No analysis paralysis. No doubt. No tracing back to origins. Just: the plan is locked, let's go.

So why can't I do this with research?

Two Root Causes

I sat with this for a while, and I think there are two root causes.

One: I keep trying to formulate my own problems from scratch. Take LLM interpretability. There exist well-defined problems online - clear boundaries, given constraints, benchmarks, exercises. But instead of picking one up and solving it, I start from some very basic formulation, then think "this is not a big deal," add some constraints, check if it actually makes sense, realize a simple solution would crack it, reformulate, and repeat. I'm spending all my energy generating and discriminating problems instead of solving them.

Two: I don't trust the plan once I have it. Even when the path is clear - say, I know I can use a few-shot scene understanding model and tune it for multi-label classification - I go back. How did these papers start? How did the architectures evolve? What are the loopholes? I know the solution exists. I know what to do. But I keep pulling at threads instead of executing.

Both of these are the same thing, really. The exploration gene is eating the exploitation gene alive.

One of my course instructors told me something recently that I keep coming back to. He said: your problems are good, your ideas are good, you know what you want to do. But don't get into a loop where you can't implement what the task actually requires. He wasn't being harsh. He was being precise. And he was right.

What the Sketchbook Knows

Now here's the part that surprised me when I noticed it.

I sketch. I scribble in my sketchbook, use different pens, draw on my iPad, post things on Instagram. I'm an amateur - I know that. But I've never once sat down to draw and thought, will I be able to draw this?

There's this idea I read somewhere: you don't need to replicate the exact thing. You need to get the essence and express it in your own style. And that's exactly what I do. I've never worried about whether my sketch is perfect or precise like an expert's work. I just look at it and think - that has something. That's not too bad. I could express my thought there.

No rabbit hole. No origin story. No tracing back to the history of the technique. Just: see it, feel it, draw it.

So I started looking at this pattern across my life. Cooking, sketching, visiting places, buying books - in all the creative, artistic parts of my life, the cycle from trigger to execution is fast. The artistic brain just moves.

The logical brain? It gets too logical. Too skeptical. Too perfectionistic. It keeps exploring when it should be exploiting. It keeps asking "is this the right approach?" when the approach is already clear.

The Real Thing Underneath

I think what's actually happening is some implicit fear. Fear of failing. Fear of the solution being naive. Fear of committing to a path and finding out it was wrong. The over-exploration isn't really curiosity - it's protection. If I never lock the plan, I never have to face the possibility that the plan wasn't good enough.

But in art, I've already solved this. I gave myself permission to be imperfect. I gave myself permission to express the essence rather than chase precision. And it freed me to actually produce.

So the insight is simple, maybe embarrassingly simple: my artistic brain should help my logical brain get things done.

Not as a metaphor. As a practice. The same way I approach a blank sketchbook page - with confidence that something will emerge, without needing the full picture before the first stroke - that's how I need to approach a research problem. Get the essence. Express it in my own style. Ship it.

The plan is locked. Let's go cook.