This is interesting because I have transitioned to where I use SOT models.. but I kind of use them like employees that I can delegate to. I still review code.

However, I now literally say.. "Here is my objective and here is a starting point for documentation. Research this and build up a plan."

This can be very company specific, like migration from one framework to another in house infrastructure framework. I'm spending my time figuring out how the plan should be chopped so I can have confidence in the parts and not overwhelmed. I don't want a tool, I want a model that can stitch resources together into a plan. That type of model is in a whole other ballpark.

The useful training data is when you clarify your intent, when you tell the model a different approach would be better, when you consistently refactor towards Y and away from X, and so on. The training data isn’t the code, it’s the session transcript. (Anthropic would call this a “distillation attack” against their model, but in this case the model is you!)

I would imagine your interactions with it are more important than the output.

Funny. I use it through Opencode Go which gives more use than I can use, but didn't realize it was actually free on Zen. Will switch to that I guess

Every lab trains their models with AI generated code at this point.

Training on ai generated content is how the models got a big jump in capability