When I started my thesis, I imagined research as a fairly open process.
You explore an idea, experiment, follow interesting directions, and gradually refine something useful.
Some of that was true.
But much more of the process revolved around something else:
Defensibility.
Not whether the idea was interesting, but whether it was grounded, narrow enough, properly framed, and embedded into existing work.
That difference took me a while to fully understand.
Why defensibility matters
The frustrating part is that academia is not wrong about this.
Research should be rigorous.
Claims should be grounded. Work should connect to prior literature. Contributions should be clearly scoped and defensible.
Without that, you quickly drift into vague ideas and unverifiable claims.
I understand that much better now than I did at the beginning.
But I also underestimated how much of the process revolves around satisfying those constraints.
A surprising amount of time goes into:
- framing
- positioning
- narrowing
- citations
- structure
- defending why the work matters
instead of actually building or exploring the idea itself.
That does not make the process meaningless.
But it changes the nature of the work significantly.
Interesting ideas are easy
One thing I underestimated is how easy it is to find interesting ideas.
Especially in AI right now.
There are endless things you could explore.
The difficult part is turning one of those ideas into something narrow, coherent, and defensible enough to survive academic scrutiny.
That usually means reducing ambition, not expanding it.
Cutting scope. Removing ideas. Narrowing claims.
That was probably the biggest shift for me during the thesis.
What I would do differently
If I started again, I would narrow the contribution much earlier.
I spent too much time trying to make the thesis broader, more comprehensive, or more generally applicable.
In retrospect, the stronger version was the narrower one.
I would also trust the contribution earlier.
Even now, after handing it in, I still struggle a bit to see the work as particularly “strong” or novel. From the inside, it often just feels like the obvious next step.
But that is partly because you spend months living inside the problem.
The work starts feeling smaller the more familiar it becomes.
Meanwhile, the actual challenge is usually not inventing something revolutionary.
It is building something coherent enough to hold together under scrutiny.
The part that felt most real
The most satisfying part of the thesis was never formatting, framing, or writing.
It was implementation.
The moments where the framework actually worked. Where the system behaved coherently across different case studies. Where the ideas stopped being diagrams and became something tangible.
That was the part that felt most natural to me.
And probably the clearest confirmation that I am more interested in building applied systems than staying in academic research long-term.
What the thesis clarified
Before starting the thesis, I was already unsure whether I wanted to stay in academia.
Now I am fairly certain I do not.
At least not in its current form.
That is not because I think research is useless. Quite the opposite.
I respect the process more now than I did before.
But I also understand how much patience it requires for process, formalism, and defensibility.
And I realized that I am much more energized by building, iteration, and applied engineering work than by academic structure itself.
Maybe that changes in the future with a topic I care about more deeply.
But right now, I cannot imagine continuing directly into something like a PhD.
Closing
The thesis changed how I think about research.
Not because it killed the interesting parts.
But because it showed me how much work is required to turn an interesting idea into defensible research.
And those are not the same thing.
