Checked October 7 and 8, 2026
Sources
How-tos that take the mystery out of building with AI, cautions worth knowing, and what research says about learning to code. This page is the background reading.
Here to build? Start on Home Hosting? Start on Hosts
How to read a card
- The title opens the source itself, in a new tab.
- What it says opens our two-line summary, here.
- For the session or The habit is what it means for you. "The session" is the two-hour session, whoever runs it; these notes were gathered for its first run, {record.ran}, and apply whenever you build.
- A topic under a card shows every source on that topic.
The short version
What coding builds. Learning to code reliably builds coding skill, and a large meta-analysis found moderate links to creative thinking, math, and metacognition, though those links shrink when compared with other active teaching.
How to begin. The research is stronger here: run something that works, change it, then make your own. Early studies of AI and beginners suggest AI help need not hurt learning when people still change code by hand and read what the AI made.
The risks. Over-reliance, confident mistakes, and data going places it shouldn't. Each has a simple habit that answers it, under Proceed with care.
How the session is built on this
The session follows a path researchers call Use-Modify-Create (Lee and colleagues, 2011), later developed into PRIMM: predict, run, investigate, modify, make (Sentance and colleagues, 2019). You start with a starter that works, guess what it does, run it, look inside, change it by hand, remix it with AI, and then invent your own. Brennan and Resnick (2012) name "reusing and remixing" as a core practice of computational thinking, and their suggestions for assessing it lead straight into the conversation about assessment that followed the first run.
Start here
A few sources in reading order, picked for where you're coming from.