It has been a while since I finished my PhD. I wanted to write down a few things I have been thinking about—not as universal advice, but as notes for people who might recognize themselves in this path.

1. Start. Try your best to start.

The first project is always the hardest. Use the resources you have and reach out to anyone who can help you. Learn, explore, study as much as you can, and try to find an entry point.

Remember that a rough start is a good start. This is always true. Interestingly, though, the way someone starts can already reveal a lot about their future research trajectory—their interests, style, taste, and judgment. Even when the work itself is rough, the choices you make at the beginning often say something about the kind of researcher you will become.

2. Be broad.

In my opinion, there are a lot of low-hanging fruits in almost every field, but sometimes experts fail to notice them because they are too narrow.

When you explore many different fields, it becomes much easier to identify interesting entry points. You can connect ideas from different areas, do a lot of “A + B,” and produce fairly solid work that can often be accepted by top conferences.

Breadth gives you the ability to see connections that people deeply embedded in a single field may overlook.

3. Be paced.

I used to be crazy about research and work. They were almost the entirety of my focus. Eventually, I realized that this was hurting both my cognitive capabilities and my overall health.

I would suggest prioritizing your overall happiness above everything else. The earlier you realize that this is ultimately the goal, the earlier you get onto the right track in life.

In other words, go inward and connect with yourself. Changing your environment or external circumstances is rarely the ultimate answer. Achieving something will not automatically make you happy.

4. Learn and grow with collaborators.

The more high-quality collaborators you have, the better your research will become.

This is difficult at the beginning. When you first start doing research, you may only have access to your advisors and labmates. Hopefully, through internships, conferences, and random networking opportunities, you gradually meet more collaborators.

Be genuinely open-minded. Working with different people exposes you to different ways of thinking, and interacting with new collaborators can help you maintain that openness.

5. Get resources.

People and collaborators are enormous resources. So are internship opportunities, APIs, GPUs, infrastructure, and many other things.

I have always tried to put myself in places where I can have more resources and more flexibility. And yes, flexibility itself, in my opinion, is an extremely valuable resource.

In some sense:

Success = people + resources + flexibility.

6. Breadth first, depth second.

In my personal opinion, breadth comes first, and depth comes second.

When I look at researchers I admire, I notice that many of them spent some time working on things that, in retrospect, were not that important. Their trajectories were not perfectly clean. They explored, tried different problems, and sometimes worked on things that did not end up defining them.

Then, at some point, they found the right track.

They found an entry point—one problem, one direction, one perspective, or one domain—that connected with the broader context of the world. They gave it a serious shot, went much deeper than everyone else, and impressed the world.

In a nutshell, breath gives you the context; depth gives you an opportunity to impress the world.

7. Depth, experience, profile, and the ultimate goal.

When it comes to an industry job, depth starts to matter much more. You may need to concentrate most of your attention on a single direction. In that environment, being extremely broad can sometimes become a distraction.

You probably should not constantly be distracted by the breadth of the research community. Instead, focus deeply on one direction and build your own moat—a body of expertise, context, and accumulated understanding that is difficult for others to replicate quickly.

I used to call this experience.

But increasingly, I feel that experience itself may matter less than it used to, because AI can give you information, explain unfamiliar topics, and teach you almost anything remarkably quickly. AI makes it easy for another smart person, someone who learns and moves very quickly, to replace you.

What may matter more is your profile: the body of work, reputation, relationships, positioning, and identity that other people associate with you.

And that is a very human thing.

To some extent, I think it is a little silly. But this is how society works.

Nevertheless, building a profile is not the ultimate goal.

The ultimate goal is still to do good research.

You want to make people aware of something they have missed. You want to reveal a perspective they have not seen before. You want to make something work much better, or make an idea substantially more effective and useful.

8. A personal reflection on my PhD.

When we are young researchers, we should have the ability—and maybe even the appetite—to absorb new things as they appear. New ideas, new problems, new methods, new fields, new trends. We should consume as much as we can, digest as much as we can, and try to understand how everything connects.

I genuinely enjoy exploring. And I still believe that this instinct is good not only for research, but also for life. Exploration gives you more perspectives, more possibilities, and more ways of understanding the world.

Sometimes I notice that truly exceptional researchers seem to complete this breadth-to-depth process very early. They explore quickly, understand the landscape quickly, find their direction quickly, and then go deep. In some sense, the earlier someone can complete this transition, the more extraordinary they seem to me.