You could also explore approaches like what puppeteer-heap-snapshot does and extract the data you're interested in from the in memory Javascript objects that the canvas powered website uses to generate its contents.
Montessori vs k5... go! But actually what about skipping k5 and starting public school at grade 2 (unsure of what age that is for the non US school systems)
EDIT: I forgot daycare is prohibitively expensive. Whoops.
Apparently all US states require 6 year olds to enroll in first grade. (homeschooling might be an exception) IIRC some of the nordic countries start school at age 7, but they require you to already know how to read and write before enrolling.
See integrating LLMs with knowledge stores a la RDF triples or anything else really if you leverage something like LLamaIndex/LangChain.
"If I were 26 in the 90s I'd be well off". Also don't necessarily worry about your career ending because "AI". Keep er movin. (And knowledge hoarding sounds gross)
maybe check out PICO 8. It's basically a VM (self styled fantasy game console) that runs on a bunch of devices and an active community. https://www.lexaloffle.com/pico-8.php
A similar project that I'm eager to hack on (or maybe try my hand at implementing something similar in WASM) is the UXN platform. https://100r.co/site/uxn.html
And if we're already talking VM platforms to explore, there's no reason that someone who doesn't want to do software professionally can't get a lot of enjoyment from some basic syntax knowledge and making modifications to software they use every day (whether that's writing a userscript in your browser or exploring your favorite game via mods)
Including less-than-ideal data in a broader analysis is completely fine as long as you're aware of the data quality issues and adjust your weights accordingly. IMO Alexander does a pretty good job of this - I recall many occasions where his writing references a study while also caveating that reference because of a limited sample size or a slightly implausible effect size or some other qualm.
There's a world of difference between trying to make sense of a topic when none of the available data is all that great, and outright fabricating data to make a name for yourself.
If you're an auto-didact my advice to myself would be: take a week to get a grip on the basics again (maybe read an algo and datastructure book or cracking the coding interview). Then grind leetcode. That's the service you're looking for : ) Or toptal.
