Hello again and Happy Holidays! For the first post of 2026, Stone Tools drafts a poor C64 into the AI wars, exploring 1985’s “expert system” software, XPER.
To learn the software, I attempt to build a weather predictor which answers the simple question, “Will it rain tomorrow?” Find out how it compared vs. my own human intuition and vs. a basic coin flip. Sputnik, Campbell’s Soup, Gary Kildall, and the Connection Machine all make appearances; I swear there is method to the madness.
Bonus: I dive into Japan’s “Fifth Generation Systems” strategy and the West’s reaction to it.
Extra bonus: a story about a guy who might have been the first person to concede his job to “AI.”
I now have the ability to add new tags, so I’ve taken the liberty of adding the stonetools tag starting with this post. Thanks always to the community for the support and reading; I had a great time hearing everyone’s HyperCard stories.
I hope you enjoy this look at a pretty esoteric corner of the C64 library.
If you use historical weather data off the net, I would like to know how well it works some day.
Feb 4,2010 predict rain 25% real weather just cloudy with meatballs.
I do recall the time when expert systems were the thing, and even the idea that an expert’s expertise might somehow be captured in a surprisingly small number of rules.
Perhaps an 8 bit expert system is mainly there to help learn the mechanics and give an idea of the possibilities?
A conversational chatbot trained on casual Q&A pairs. Responds to greetings, questions about itself, and general banter with terse personality-driven answers.
> hello
HI
> are you a robot
YES
> do you dream
MAYBE
I’m not convinced the 8-bit version was just an exploratory “learn the mechanics” project. XPER was built for a specific biological task, adopted by a fair number of scientific communities wishing to catalog various taxonomies, and cited in multiple scientific papers.
Now, in marketing it as a general purpose expert system, it is entirely possible that the creator then learned how much he still needed to learn about building such software. But he did absolutely intend for it to be used for legitimate scientific research, as evidened by the fact that XPER2 and XPER3 were developed and are still available and in use today.