Meetings are for technical progress and project planning. On the technical side, we work together to chip away at tough problems, help everyone get unstuck, and figure out immediate next steps. On the planning side, we set bigger goals, track progress, and adapt as we learn.
Meetings are important work; they should never distract from getting stuff done. We will only meet when we have a plan and concrete goals, otherwise we can just hang out in the lab. We will end meetings early if we can accomplish our goals quickly, or cancel them outright if they do not have a good plan 📕. Doing so provides more time to work in the lab together and just enjoy life.
Toward these goals, we will collectively maintain regular meeting cadence and structure.
The beginning of the week is tactical, focusing concretely on what we want to accomplish in the next few days. The end of the week is strategic, reflecting on progress and how we should adjust plans.
Mondays are for all project meetings. Every student should go to one project meeting1. It is better to do one project exceedingly well rather than several projects at a mediocre level. Requiring all project meetings to be on Mondays also serves as a sort of “budget” or limiting factor to ensure that I can regularly invest good hours helping on your project by avoiding overcommitment; please hold me accountable if you see me violating this rule.
On Tuesdays I will work in the lab. I hope you will join me! Historically, working together in the lab is where the best meetings happen because they are motivated by an immediate problem that we can quickly chip away at together. If you see me in the lab, I am there to collaborate, so don’t hesitate to “interrupt” me; that’s why I am there.
During the school year, we will have group lunch at 12:00 in the PLSE Lab (Gates 253) with research presentations starting at 12:30. To the extent possible, all our exams, practice talks, and visitor presentations should be scheduled on Tuesdays as a PLSE Talk. However, my favorite PLSE Talks are work-in-progress discussions and live demos of prototype tools where everyone in the lab can help brainstorm.
Tuesdays afternoons are also available for quick one-off meetings, just message me.
Wednesdays are for doing big things. Avoid scheduling any recurring meetings on Wednesdays. Coordinating our “no meetings day” helps everyone have enough time to do deep, technical work. Wednesdays are a good time to (on demand) find a conference room and actively pair program / write / prove with your team for a couple hours.
If for some reason we cannot schedule an exam or practice talk during a Tuesday PLSE Talk slot, then Wednesday afternoon is an acceptable (though not ideal!) backup option.
On Thursdays I will work in the lab. I hope you will join me! Historically, working together in the lab is where the best meetings happen because they are motivated by an immediate problem that we can quickly chip away at together. If you see me in the lab, I am there to collaborate, so don’t hesitate to “interrupt” me; that’s why I am there.
The Allen School Colloquium will be at 15:30. Please participate! In Autumn, colloquium will be a mix of distinguished lectures and talks from labs across the Allen School. In Winter and Spring, colloquium will mostly consist of faculty candidate talks. These talks survey what’s going on across the broader computing landscape. Even better, they are a great opportunity to observe what works (or doesn’t) when giving presentations. To get the most out of colloquia talks, consider trying the “Three Things” exercise!
Thursday afternoons may also be OK for quick one-off meetings. Again, just message me.
Fridays are for:
30 minute one-on-one PhD student meetings with me in the morning
Planning our goal for Monday’s upcoming project meeting
30 minute team-wide progress meeting at 15:00
60 minute PL Reading Group at 15:30
After our one-on-one, you should prepare an agenda for our upcoming project meeting on Monday. See below for some ideas on what makes for a good agenda.
At 15:00 we will have a 30 minute progress meeting. The first 10 minutes of the meeting will be for folks to make a single slide highlighting their progress from this week. You should also make another slide with a screenshot of our shared project notes doc with the goal and agenda for our upcoming Monday project meeting (this means the agenda has to already be in our shared doc before the Friday Progress meeting!). The next 20 minutes of the meeting will be for us all to go through the slides together and hear what everyone has been up to.
For your progress meeting presentation slides, remember that you have an audience. Part of the goal of this meeting is for us all to better understand what everyone is up to, that way we can better support and encourage each other. You will never have enough time in one week for folks to grok what you’re up to, but week-over-week the big picture should start to gel, especially when paired with occasional PLSE Talks etc.
For the progress meeting, you should already have the agenda for the next Monday project prepared before 15:00, and and it should be written out in our project meeting notes doc. You are encouraged to also prepare your personal progress slide before the meeting, but it is 100% OK to make it during the first 10 minutes of the progress meeting; that is why that time is reserved. We will make a fresh deck each week in this shared folder.
At 15:30 we will have PL Reading Group (PLRG) to end the week with a lively discussion. Please check out the PLRG Principles and the PLRG Schedule.
The quarter system is short and intense. On week 11, we will cancel all regular meetings and push hard to wrap up our quarter goals. On week 12, we will cancel all regular meetings, rest, and plan out the next quarter.
Structuring frequent tasks is useful because it frees up thinking and emotional energy for other tasks. Below is a default meeting formula we can adapt as our team and projects evolve.
Each meeting should have a:
Ideal meeting room reservations:
|Gates 253 (PLSE Lab)
|Gates 201 (Zach’s Office)
Every meeting should have a concrete, specific goal written out explicitly in the meeting doc. Think “meeting in a sentence”. Everyone should know the goal a day (or more) before the meeting.
Here’s an example of a simple and perfectly good meeting agenda:
Draft pseudocode for main algorithm
Here’s an example of a more detailed, but also good meeting agenda (though maybe too ambitious):
Settle on operators for first DSL prototype
- Discuss tradeoffs for restricting only to affine cases
Pros: easier to get end-to-end prototype working
Cons: other components might overfit to these cases
- Agree on specs for each operator
Pick three example benchmarks to implement this week
Debug performance regression from nightly eval run last week
See logs here (link)
Devise strategy to reproduce
- Not observing this on my laptop or in CI
Good agendas are:
Concise. Folks can quickly determine how to prepare.
Simple. Folks can easily determine what’s needed for the meeting to accomplish its goal.
Realistic. The goal is achievable within the allotted meeting time.
Prioritized. The goal advances the most important project tasks.
In contrast, here is an example of a BAD meeting agenda:
Look at code
Start on time and stay on track to accomplish the goal before the meeting ends. This can be very tricky when each cool idea sparks three more cool ideas. Just jot such ideas down; we will explore them in the lab together after the meeting.
Keep updates short. Ideally updates will largely happen in one-on-one advisor meetings, progress meetings, and opportunistically in the lab. Most meetings will still need to begin with a few minutes of updates to get everyone synced. Watch for any updates that should modify the meeting goal and be ready to adapt.
Ensure everyone participates. This can also be very tricky. I will do my best to watch for any issues and help handle them early. If you notice an issue that isn’t being handled, please put it on our one-on-one agenda or schedule a one-off to discuss.
Reserve enough time to wrap up with concrete tasks for the week. Think “What do I want to say I accomplished by this Friday?”
One-on-one advisor meetings are for regularly checking in on how you are doing, long-term (post-PhD) goals, timing for TAing and internships, etc. We should work on whatever is most important for your happiness and long-term success. That can also mean covering the same sort of topics as project meetings. If you are mentoring another student, we should always make time to briefly check in on their progress as well.
Like project meetings, it’s often helpful for our one-on-one meetings to have an explicit goal. Unlike project meetings, we should usually still meet even if there is no explicit goal. Checking in and making sure things are going well is always a good goal for one-on-one meetings.
Folks can do roughly four hours of challenging technical work on their main project each day. Since “peak working hours” are precious, planning how to make the most of them is an excellent investment. Having explicit structure frees up time and energy for more interesting work and helps manage expectations for everyone in the group.
We have myriad other responsibilities on top of research. The challenge is that all these other things are important too:
These all help support “the point” of a PhD, i.e., developing the skills needed to conduct original research, but they often contribute somewhat less directly than just actually doing research. Avoid the trap of death by 1,000 papercuts.
It’s very easy to get lost among our many obligations and procrastinate on the challenging, important tasks that push our research forward. Pulling “stunts” to cram a week’s worth of work in before our next meeting or attempting back-to-back all nighters for a deadline is very bad. Not only does it lead to lower quality work, but it will destroy your health over time, both physically and mentally. It is much better to establish and maintain a steady, sustainable routine.
Having too many meetings is a trap. If we are so busy we can’t prepare for or be fully present during a meeting, then the meeting just drains energy and eats into the good working hours it was supposed to support! It is better to just cancel such meetings and get caught up instead. Of course, the best practice is to never get so overcommitted in the first place.
Hopefully the ideas in this document will help. Of course, this is all just a big experiment. Please message me if you have any suggestions2. We will try it out, iterate, and adapt as we learn!
10 Bullets by Tom Sachs
Make Plans Shorter by Lucas F. Costa
Speed Matters by James Somers
Avoid the XY Problem
Maker’s Schedule, Manager’s Schedule by Paul Graham
📝 notes index
Maybe two. Again, focus is key, avoid overcommitting.↩︎︎
Many thanks to the following folks for feedback on this doc: Anjali Pal, Chandrakana Nandi, Vishal Canumalla, James Wilcox, Dan Grossman, Amy Zhu, Pavel Panchekha, Steven Lyubomirsky, Adriana Schulz, Max Willsey, Chris Thachuk, Yihong Zhang↩︎︎