Meeting with Wayne and Steve. We were talking about some upcoming interface and application changes. Lot of talk about using AI to change things. There are some pros and cons. Here are some of my notes:

- We are burning through tokens.

- Skills and teaching the AI what we really want (rules and processes). We are seeing some problems with consistency and some other issues.

- AI's default fix is to put a guard around it (if exists, do this, if not, do this).

- No real way to test our code.

- AI is taking our code and creating different versions of key features. Multiple implementations for the same thing (ish). We would love it to use our existing processes.

- Making consistent changes - that's what we are wanting. Slight wandering and varying of the code source.

- Back to testing and setting up testing accounts. We are starting to get into automating some of that. Still needs some help.

- Steve is heavy into AI and that is how he is coding everything.

- He is loading in things to the demo sites. Eventually, he would even like AI doing demos, AI doing payroll updates, etc.

- Cory is doing a little bit of backend code stuff. She gets the call from the customer and then she makes the fixes using AI.

- Wayne wants to make the AI more productive and more standardized.

- We have pages and functionality expanding at a crazy rate. We are also multiplying possible problems. Being able to manage all of the different possible configuration options.

- Making customers mad over time due to so many changes.

- Test driven design skills - templates and how to apply them over time. Helping to remove the need to look for AI context. Send it to a specific template and then let it run, based on those templates. Build in the patterns and context.

- Talking about "View Objects" and splitting things into MVC (model, view, controller) stuff. Separating the layers - the business logic, the view layer, etc.

- Wayne is working on some workbench stuff - adding in fixtures and scenarios to run against certain pages. Being able to switch themes. The view object becomes the record of truth. Checking the visual impact of the different changes. Converting some of our older pages into testable views. Trying to bring it (all of adilas) under test, without making major changes.

- Pushing all of the data into a valid view object.

- Setting up new pages that follow our new rules, processes, with view objects that are testable.

- Focusing on the main traffic - pages that get tons of views and traffic - like view cart, etc. Wayne wants us to pick some pages and go slow, vs just telling AI to run with it. Trying to get some consistency up and going.

- Everybody is working on work in progress stuff. Be patient. We may be giving AI too much access and authority.

- Being able to really translate the background meaning of things (intent, assumptions, and overarching context). Coordinating all of the moving pieces.

- If we want to customize things, help make it fast, powerful, and stable. Custom is part of our model.

- Wayne spends lots of time fixing new AI code and missing variables (he said 75% of time over the past few months). It works great but we need those guardrails. He is on a different time zone then we are... New things will launch for us over night. It is morning time for him and his error logs start going off. He ends up changing and fixing things before everyone else comes to work. It is almost turning into his daily process. AI is super fast, but we may be introducing errors and bugs even faster. I'm so grateful that Wayne has some logging and error monitoring tools in place. That is huge.