Recording notes from the last couple of days (11/13/25 to 11/15/25).

Emails and sending an email to Thomas with some user/payee id numbers for adilas super users. These will be used to help train the AI agent.

The Nxtlinq guys are adding in more logs so that we can track some other things. They are working on a number of things including improved memory and global context stuff. Memory in AI is making training and knowledge stick. We also talked about a number of other topics. Here are a few of them:

- The costs for the infostructure and keeping the AI agent up and running.

- Talking about global information storage - we really need this, from our team to the bot.

- Talking about influencing things upstream and downstream. We need the AI bot to remember and learn, but we also have to be careful about where it gets its information.

- Light talks about an audio issue between multiple tabs. The AI agent is still listening in a different tab and responding. The simple work around is the mute button on the microphone. We may look deeper into this, if it becomes an issue.

- After the main meeting, Steve, Thomas, and I stayed on the meeting to talk about pricing and future plans.

- Being able to train the agent so that it can help with support, setup, and deployment. Figuring out the pain points and helping us as a company as well as our clients and users.

- Thomas was telling us some stories about introducing agents for other companies.

- Steve was talking about being able to help get the agent up and going and helping with oversight.

- Lots of talk about ingestion rates and values (how much we feed it).

- Human in the loop to help with the training.

- Thomas talking about manual crank windows (roll up/down the windows by hand) vs electric windows (power windows). Ideally, we show our clients and users how awesome it is with AI help. Then, hopefully, they won't want to go back to older tech.

- Getting into the BI level (business intelligence level) later on, as we go.

- If we train this agent up, it will really help our users and clients with training, consulting, etc.

- Here are our steps and phases for working with the AI agent - 1. Navigation and prompts, 2. Setup and training, and then 3. Clear out to consulting.

- Small phone call with Steve afterwards to go over our internal plan a little bit. Good meeting.

Meeting with Hamid. Showing him some graphics and then playing with the adilas AI agent inside of adilas. One of the goals for the AI agent is... training and retaining (keeping or holding the info). We then worked on getting his VS Code instance up to date (his code editor). We played around with some code and letting Copilot (a different AI bot) help with some of the coding. We also talked about the value of knowing what is going on and/or what is being prompted/generated by the AI bot. Especially in code, the bot may give you a bunch of code, but you don't know what it does and/or how to fix it or tweak it slightly. You can get yourself in trouble. You still have to know code. Fun little session. Lots of show and tell and light playing around with different tools.

Work session with Shannon. She really helps me carve out time to work on some of my long term goals. If it weren't for her, I would get sucked into the main mix and other demands inside of adilas. I love it that we get to work on these things (other projects and dreams). As far as work today, we spent most of our time working on new content for the fracture buildout section of the adilas lite plan. Today we were finishing up the stuff on front-end and backend validation processes. Here is what we came up with:

//////

This validation section deals with security of the data being passed around, and/or transferred, from client to server and back again. The term client-side validation deals with checks and balances before the webpage or form is submitted. This front end, or client-side, validation can really help with the user interface and user experience (UI/UX) parts of the system. This type of validation can help catch simple mistakes, typos, or missing information before the data is sent to the server. This helps speed up the process and reduces server load. If the data doesn't pass the client-side validation checks, then the user is prompted to correct the errors before submission. This is a great way to enhance the user experience and reduce unnecessary server requests.

Server-side, or backend validation, deals with what goes on once the webpage or form leaves the user's computer and hits the server. The server is the backend brains of the application. Backend validation is responsible for checking and making sure that anything submitted is viable and can be used in the requested action. This backend validation is very secure, it can exist in multiple levels, and may even be stacked if needed. If backend validation is used, and the data doesn't pass the validation checks, then the server throws an error. Depending on how deep it is, the errors will stop the process and the desired action is not completed. One or more error messages then get passed back to the client. Error messages are required for communication with the users. However, they may lightly detract from the user experience.

The best combination is a good mix of both client-side (front end) and server-side (backend) validation. We, as a company, are already doing a lot of this mixed validation. We are currently heavier on the server-side, but both options are being used. Moving forward in the fracture buildout, we will continue to keep the server-side tight. We also plan to enhance the client-side validation to make the whole experience even smoother for our users and clients.