My kids both have ADHD. They both have trouble remembering things and can get easily distracted. When I grew up, we had notepads and could easily write things down. Nowadays, we tend to rely on technology, and while I still default back to using a notepad for many things, it’s just as easy to “set it and forget it” in my calendar. For as long as I have been in the web development and IT industry, I have had a smartphone (BlackBerry alum here!) that basically ALWAYS keeps track of my calendar and appointments. Kids don’t always have that though. Or, there are too many distractions and games on their devices that make it impossible for them to pay attention long enough to actually put that information into their phone.
How many times have I said, “you need to write this down” ? More than I can count.
I started thinking about how I’ve been applying artificial intelligence in my web development sphere. It accelerates the development of custom code, including plugins and specific web applications that we develop. While I was travelling to a family event, I was speaking to ChatGPT and hashing out the details of what a memory-based webapp that integrates with AI and has safeguards and guardrails for youth would look like.
Most of my thinking was from a technical standpoint, that it was how to handle the data, how to work with the database, and at what point did the AI get involved to offer oversight and analysis of the content that was coming in and out of the chat. We discussed things like using it as a communication and memory aid only, not an academic assistant – so it wouldn’t be able to help with homework, answer questions, or do anything other than help them as an external brain.
The workflow ended up using a proxy that would have basic parsing capabilities, and be able to detect if the question already existed in memory. I used regex to handle this type of task. If it wasn’t able to figure it out, then it would send it to AI with context. The context would be parsed from whatever the chat text was, using common phrases and specific sentence structure. If the AI detected that it needed to store the information as a memory, then it would save it to the database and notify the front end user that it had saved the information.
The goal for this was to add a parent caregiver dashboard, but continue to maintain the privacy aspect of it; with no personally identifiable information being stored, only an anonymized identification code. I planned to use session tokens such as cookies and PHP session memory storage to reconnect a user to stored memories. The other side of it was for youth who had learned that one parent could be counted on to give in to their demands, so they learned (either consciously or subconsciously) that they could triangulate, and get what they wanted. Or for youth in care, if they had a particularly challenging set of circumstances – where having an external brain there to remember things and keep track of “who said what” would be a welcome anxiety relief.
After planning out the details, and the technical side of things, I turned to Claude to help flesh out the framework of it. I utilize several tools, as each does specific things better, but Claude for setting up a new project is incredible.
