The most useful person in any planning session is the one who asks "but what if you're wrong?"
I kept finding that when I was excited about an idea, I'd talk myself into it. The reasoning would start with the conclusion and work backwards. I needed something that would push back before I committed.
So I built a mentor system. I called it Solomon.
What it does
Solomon has a few modes. The most useful is adversarial audit — you describe a plan and it actively tries to break it. Not to be difficult, but because a plan that survives a serious challenge is more reliable than one that was never challenged. It looks for race conditions, missing error paths, assumptions baked into the design, and places where a partial failure would leave things in an undefined state.
It also has a debrief mode. After something notable happens — a trade, a deployment, a build that took longer than expected — you describe what happened and Solomon asks what you learned and what you'd do differently. The point isn't to record the event. It's to extract the lesson before the next thing takes over your attention.
The continuous learning hook
Every interaction with Solomon emits to a continuous learning system. If Solomon raised a concern and I changed my plan, that's recorded. If I pushed back with good reasoning and didn't change the plan, that's recorded too. Over time, the system builds a picture of which types of challenges are consistently useful and which are noise — and adjusts how aggressively it challenges based on that history.
The challenge I accepted today becomes the instinct Solomon proactively raises tomorrow.
I've been running it for three weeks. The most useful thing it's done so far is point out twice that I was solving the wrong problem — once in a trading strategy design and once in an infrastructure build. Both times I was convinced I knew what I was doing. Both times it was right.