During my learning sessions, I told Claude to store all lessons, rules and standards that stood out. I never specified where to save them, and I never really checked them, because asking Claude was always faster.
It generated 15 files that held the answers I kept asking Claude for. So, I designed a new system that changes the goal from only storing to also recalling. Now Claude turns the same information into open-ended questions.
I answer 3 of those questions cold every morning (without looking anything up), a whole topic every Saturday, and full recall at the end of the month.
Three days after I “understood” pass-by-value in Java, the morning question was: a method gets a dog named “Rex”, renames it “Max”, then swaps in a brand-new dog. What prints? I said “Rex”, because I thought the method couldn’t touch the caller’s object. It printed “Max”. A method can’t change which object you point to, but it can change the object itself.
As a QA, I was testing my own knowledge, the same way I test software: outside the conditions it was built in.
The system showed me the gap between what I thought I knew and what I actually knew. This question appeared several more times in different forms, and eventually I got it right, with the full explanation.
The biggest test came in October. Due to personal reasons I put my learning on for 2 weeks, and I wanted to see how much I still remembered when I got back.
Two-thirds were still there: 19 questions were fully right, 9 were partially right (meaning some details were missing), and only 1 was missed. The rest told me exactly what to re-learn, and 4 days later 9 out of the 10 came back fully right, and one was partially right.
The 15 files that I never opened became questions that I answer every day. What’s the last note you wrote that you actually opened again?