Thoughts

Quick thoughts and observations.

Sometimes education research can be so frustrating. Student school-to-school and district-to-district mobility is critical to understand achievement and interventions. There is no public measure of this readily available at any level of aggregation. Right?

Thinking about going back to posting a picture a day. For awhile it was a good ritual to keep me connected to my photography even when I wasn’t shooting, and to help me be a little more reflective about what pictures worked or didn’t and curate my own taste more. I need to do the same with my writing! But one step at a time!

“Our chief weapons will be better schools, and better health, and better homes, and better training, and better job opportunities to help more Americans, especially young Americans, escape from squalor and misery and unemployment.” ~ Lyndon B. Johnson (from Caro’s biography)

I love riding my bike in Boston to get around. But it is radicalizing. You encounter a hundred little frictions and you see why infrastructure is hard. And you pine for universal design. A city that works for all of us is better for all of us!

Spent some time today pulling all my “private” code repositories from GitHub and self-archiving them on my self-hosted Gitea instance. I should have done it awhile ago - private is no longer private on GitHub. Feels good to continue the liberation.

The new Olivia Rodrigo album out today, you seem pretty sad for a girl so in love, is really good. I’m not sure I’m the audience she’s going for so I hope it still does well - but it is just a great listen, perfect for a summer drive.

I am in sports watching nirvana with a great NBA finals, the World Cup starting, and the Stanley Cup Finals with two teams I don’t love, but playing at an incredible pace.

New website is live and I am so happy about it! Lots of work to get here, but the post once, syndicate everywhere workflow is great, and having it independently operating on my own infrastructure feels good!

Much of my early career was peering into black boxes: using open source statistical software, demystifying ML models, communicating causal inference in plain language. With AI I am building AI-assisted tools that do the same. Openness and trust - these are still core to my work.

My first real use of AI for something productive was asking openweights models Mistral and Code Gemma (v1 I think!) for help with regular expressions. Now I built this website with Claude Opus using the Claude Code harness less than 2 years later.

Working paper verticals like EdWorkingPapers, NBER, and others are a convenient way to end-around the publication process while retaining a lot of the prestige and value of submitting to peer review.