why the posts?
1. Motivation
This blog is mainly a record of my learning process.
I am currently studying topics around large language models, but my understanding is still developing. Because of this, this blog is not meant to present polished research or definitive conclusions. Instead, it is a place where I try to learn more carefully by writing things down.
The idea is that if I can explain something clearly, then I probably understand it better than before.
2. Why Write?
A lot of technical learning is passive—reading papers, going through blogs, or coding up small experiments. While this helps, it often creates a shallow sense of understanding where things feel familiar but are not fully clear.
Writing forces a different kind of thinking:
- unclear ideas become obvious
- missing steps become visible
- assumptions have to be stated
In that sense, writing is not just for sharing ideas, but for actually forming them.
3. What I Plan to Write About
The posts here will focus on small and specific questions, usually things I don’t fully understand at first.
These could include:
- trying to re-derive a result from a paper
- comparing two methods that seem related
- understanding why something works (or doesn’t)
- combining ideas from different sources
I am not aiming to cover topics broadly. Instead, each post will try to resolve one small confusion or question. Posts could also be some experiments I want to try out (though they might have been explored to death in mainstream research)
4. Why This Approach Makes Sense (for Me)
Given my current level, it doesn’t make sense to aim for big or novel ideas immediately. But that doesn’t mean the process is not useful.
There are two reasons I think this approach is still valuable:
(i) Many details are skipped in papers
Even well-written papers often:
- skip intermediate steps
- assume familiarity with certain concepts
- leave some reasoning implicit
Trying to fill in these gaps can already be a meaningful exercise.
(ii) Connecting ideas helps understanding
Sometimes different papers or methods are closely related but not presented that way.
Trying to connect them can:
- make things easier to understand
- reveal similarities or differences
- suggest small extensions or variations
Even if nothing new comes out of it, the understanding improves.
5. How Posts Will Be Structured
Most posts will roughly follow this format:
- Question — What am I trying to understand?
- Background — What do existing sources say?
- Exploration — Working through the idea
- Takeaways — What became clearer?
- Open Questions — What is still confusing?
This is mainly to keep things organized and consistent.
6. Limitations
Since this is part of my learning process:
- some explanations might be incomplete
- some interpretations might be incorrect
- my understanding may change over time
I think this is fine, as long as the goal is to improve gradually.
7. Closing
This blog is essentially a collection of structured notes made public.
If a post helps me:
- understand something more clearly, or
- think about a problem more carefully,
then it has done its job.