Personal Kernel

A few reminders worth keeping in memory:

Don’t be a dick.

Give a shit.

Every decision is optimal.

Stay on policy. Off-policy data doesn’t matter.

Your actions affect the world.

“Optimal” does not mean every decision is good. It means that a decision is optimal under some model of the world, some objective, and some set of constraints. Before calling an action irrational, understand what it is optimizing for.

Stay on policy: learn from the trajectory you are actually creating instead of living through counterfactuals. You are not outside the system, either. What you do changes what happens next.

That is the kernel. Everything else is commentary.