I have no problem with personal.md.
What I have a problem with is the idea that AI should get my entire life as input. Finances, health, relationships, fears, goals, writing style, old patterns, the five-year vision. That is not neutral “context”. That is the frame the model will start interpreting you through.
And AI turns into a feedback loop very easily.
Describe yourself through your limitations and the model will start justifying them elegantly. Give it the wrong context in the wrong order and it will latch onto the first frame and keep building on it. Dump in too much and you burn tokens, privacy and control.
I prefer to treat context as architecture, not as a diary.
Often I have no organised brief to start with. I have a current thought, a tension, or some energy I’m trying to squeeze something out of. I write the first version and watch what the model does with it. If it goes somewhere good, I add another layer. If it goes somewhere bad, I change the prompt, the role, the assumptions, or the entire conversation context.
I don’t want AI to know all of me. I want control over which version of me it amplifies.
The best context isn’t the biggest file.
The best context is your current thought.