mastodon 4.7.2This account has moved servers to @pRobably@aus.social
✏️ I wrote a thing. Week 4 of my adventures building a teaching tool with an LLM agent.
https://headtilt.me/cat-herding-wasm-python-04/
I have a link to the app itself in this post, since I'd shared it with a few people beforehand and just not told anyone.
This week was about teacher feedback through hashed test suites, and files in testing and userspace, as well as some introspection about using Micropython in the project.
✏️ I wrote a thing. Week 3 of my adventures building a teaching tool with an LLM agent.
https://headtilt.me/cat-herding-wasm-python-03/
It’s homing in on the end of my feature list.
It’s a client-side only (WASM + JS interface) tool for giving students programming problems, an interface to letting them write, run, and test their solution, and providing interactive feedback and unit tests. Only relies on basic web hosting, no fancy infra. I think it’s pretty cool!
Part 6 of my "let's make a client-side Python teaching tool" series.
https://headtilt.me/cat-herding-wasm-python-06/
This week I added a record-replay feature (short vid in the post) where after a successful run, students can walk back through their program's execution. The UI needs to be a bit more obvious, but it's working nicely (so far!).
Opened up the repo for my little side project from the last few weeks: Clipy (pronounce or capitalise it however you want).
It’s a 100% client-side app for presenting (micro)Python problems to students requiring just a static web host. Problem authors can define feedback rules and create unit tests and students can do them, loaded up from a CORS-friendly host (like raw Github content).
https://github.com/zarify/clipy
Still a WIP, but I think it’s pretty neat so far.
Little bit of an update on my WASM Python teaching thingy (good name eh?!).
https://headtilt.me/cat-herding-wasm-python-05/
Bit of a mix of improvements to authoring (can load problem sets, define feedback rules that depend on other feedback rules) and student experience (persistent success indicators for passed test suites, nicer feedback presentation, added markdown admonitions).
Link to it in the post if you want to have a play. Meant to finish some docs this week but it got busy.
✏️ Part 2 of my "let's build a client-side Python teaching tool with agents".
In which Rob gets frustrated by many regressions but is actually getting something useful.
Later down the line when I try to figure out how much of the code can either be refactored away or just thrown away is going to be Very Funny.
✏️ I wrote a thing. It's about using agents to help shift a 'maybe someday' project off the backlog.
I've been wanting to make a WASM-based "learn to code in Python" thing for ages. Started a couple of times and hit some walls I didn't know how to get around or over.
Agents helped get a bunch of that moving. Maybe I'll actually get it done one day, but it's looking more promising!