A semester of milliblogging

I live blogged my class and lived to tell the tale.

Back in July, after a month of experimenting with substacking, I decided to try and live blog my class.

I’m teaching our graduate introduction to machine learning again this fall. For whatever reason, I can’t get myself to teach this course the same way for too long. I rewrote the syllabus last week and think I have a newly compelling story this year. I’m going to be teaching from the book Moritz and I wrote, but I want to teach the course with a critical eye toward the 2nd edition. Which parts of the book do I still agree with? On which parts have I changed my mind? I will blog lecture summaries every Tuesday and Thursday, and I hope to continue the discussion here.”

And I guess we’ll revisit where this substack is at around winter break. I appreciate all the feedback you’ve sent me on my writing thus far. Please keep me honest as this milliblogging moves forward. Let’s see where this goes.

So where did it go? The live blog experiment was one of the most rewarding pedagogical challenges I’ve attempted in my career. I’d strongly recommend you give it a go if the circumstances are correct. Let me first touch on what I learned. But then I have a bunch of questions for you, my readers, as I’m trying to make some decisions about the uncertain future of this substack.

When I set out to start substacking, I thought I’d use it instead of wasting my life on Musk’s Twitter. In the hour between when I’d wake up and go to the gym, I’d write short little blogs instead of doomscrolling. This use case is a bit problematic. Substack is designed to be a newsletter system, and emailing people every day with short little shitposts was bound to be annoying. For better and for worse (mostly for better), Substack is not livejournal. I quickly realized I had to think a bit harder about what I was sending out. The commitment quickly became more intense than a Twitter habit.

Committing to blogging a class was a whole extra level of intensity. Most days, I could write these blogs in an hour as I intended. But there were certainly days when these blogs took much longer than I wanted them to, sometimes requiring three hours to finish. I don’t think that such time commitments are sustainable. On the other hand, I usually learned a lot when I got stuck. That’s how learning works, right? Maybe periodic writer’s block is an integral part of the process. Whatever the case, let me warn you that tightly bounding the time commitment of class preparation might not be possible.

It was worth the commitment. I found myself more deeply engaged with the material and more willing to hammer out the things I didn’t understand. Trying to communicate conceptually difficult material is a unique exercise and part of why I started this substack. Graduate classes are almost inaccessible by definition. I struggled with balancing language for a broad audience and language for technical precision. The middle ground isn’t always easy to strike.

And with some practice, I did get into a deliberate rhythm for how to run a class live blog. On the day of the lecture, I’d write some opening monologue for what I hoped we would learn. This writing would let me settle my thoughts on the structure of the in-person lecture. Invariably, the lecture would go differently than I planned, and the following day, I’d write up snags or difficulties or parts that I found surprisingly compelling. An intro blog and outro blog made for a good cadence, and I think it’s a sustainable model for next time.

Next time will have to be for a different class. I’m not convinced you can live blog a class twice. The material I put up here was thinking through revisioning a text book in real time. If I taught the class again, I’d probably just try to take these blogs and turn them into a coherent full text (more on this in a minute). I also don’t think you’d want to read a live blog twice: most people only want to take a class once. Only professors are kooky enough to take the same class over and over and over again.

That said, live blogging got me to practice writing, attempt better scientific communication, and learn things I didn’t know. I found this process so valuable that I aspire to blog every class I teach at least once.

Moving forward

The comments and emails I’ve received have been the best part of this process. I value your feedback! I know the amount of text I’m generating is overwhelming, and I also know that the email delivery system of Substack isn’t as good as Google Reader. I know getting an email from me every day with 1000 words is a bit much, but I appreciate all of you who have taken the time to engage.

Along those lines, I want your opinions on some decisions about how to move forward with the Substack. First, I’m not sure what I should do with all of this text from the fall! I’ve made a table of contents so you can go and find the blogs in one place in the order they were written. The table of contents is linked at the top of the main page:

If you think it would be useful for me to add more annotation to this ToC, please let me know in the comments or by email.

I’ve had a few people suggest I try to compress these blogs into a single manuscript, like a companion text to Patterns, Predictions, and Actions, potentially written without equations. If you’d like to read such a book or think others might be interested, please tell me what you’d like to see.

I also have to figure out what to do next on Substack. Let me save that conversation for tomorrow. But if you have thoughts, please let me know before that. Comments are open, as is my email. If you send me thoughts today, I’ll address them in tomorrow’s installment.