How We Do Research
Living 中文
For students and postdocs who may join us, for people who want to collaborate with us, and for anyone who is simply curious about how we do research.
This section collects the philosophy and the working style behind how I (we) do research: what I believe, how we discuss and work day to day, what I promise students, and what I expect students to do.
You can read straight through in order, or jump directly to the part you’re most interested in.
- Six Things I Believe —
- Most ideas won’t work, but we’re not afraid to try and to debug
- Use toy examples to attack existing methods, and to give new ideas a fast debugging path
- A problem worth doing must be both important and within reach
- We prefer fewer and deeper
- Build intuitive, everyday thinking, and learn to connect different experiences
- Discuss more, express more — the quality of an idea has nothing to do with seniority
- How We Discuss
- Feedback is there to make us stronger, not to prove who’s right
- It’s fine not to know; bluffing is what doesn’t help
- Understand before judging; ask before concluding
- Newcomers, please speak up
- Harsher feedback in private; in public, mostly questions, recognition, and suggestions
- How the Day-to-Day Works — group meetings, one-on-ones, the implementation ladder, answering the questions clearly before you start, open-source code.
- Authorship and Research Contribution — actual contribution is the sole principle for author ordering.
- Promises and Expectations — what I promise you, and what I expect from you.
- Your First Month — suggestions for the early days after joining.
- For Collaborators — how cross-field collaboration works, and the timeline realities.
If you’re considering applying, this section is also what question 10 of the Research Reflection Questionnaire asks you to respond to.
Who this place may not suit
- People who want very explicit instructions at every step. I’ll give you a starting point and context, but not the detailed route.
- People who can’t work in uncertainty for three months. Research often goes long stretches with no visible progress (even if our toy examples give a little bit of joy).
- People who see admitting mistakes and accepting feedback as weakness — our entire way of operating is built on the opposite assumption.
This section will keep changing. If you think something here is wrong — or that we say something and don’t do it — tell me directly. The second case matters most.