There were a lot of great submissions for EmacsConf 2022. The only way we were going to get all of them into EmacsConf was to split it into two full tracks, which we've never done before. We hadn't planned on changing much about EmacsConf this year, but fortunately the other organizers let me talk them into going for two tracks.

I've been spending the last few weeks figuring out the tools and processes that will need in order to scale up, taking advantage of A-'s participation in virtual grade 1 to get some focused time during the day and W-'s patience to squeeze in some more coding here and there at night.

Most of the code I wrote last year needed some tweaking to handle multiple tracks. I rewrote my scheduling code to allow me to quickly test different strategies, and I learned how to use Emacs's SVG and XML support to make graphical views of the schedule. I added automated validation for availability constraints, which made it easier to shuffle talks around.

Amin Bandali usually handles the infrastructure, but he's busy, so I started digging into it as an excuse to learn Ansible along the way. I want our setup to be easier to reproduce, and configuration management helps me prevent some of the mistakes that can happen when I'm doing system administration while distracted or in a rush. The playbook is starting to look pretty reasonable.

I'm looking forward to refining our captioning workflow, too. Participants really liked watching the videos with captions last year, and I was happy to see that the people who helped make that possible also want to help out with EmacsConf 2022. I've started trying out OpenAI Whisper for captions and lhotse and torchaudio for word alignment. Recognizing text with the large model takes a long time on res.emacsconf.org because it doesn't have a GPU, but the results are pretty nice.

Looking forward to learning even more in the lead-up to the conference!