Inspired by Brett's and others' commentary lately, I put my old equity research hat on (I was on a top II-ranked team at Evercore) and built a financial model for a packaged food company I know extremely well... using all of my best practices, knowledge, and dream wish list features. The process took about 4 hours. 2-3 hours of that was Opus 5.5 grinding out compute. I burned a noticeable chunk of my AI usage credits in the process. It is an absolute machine. A work of art. It's literally the greatest financial model I have ever seen. Highly unlikely, dare I say zero chance, that anyone on Wall Street -- sell side or buy side -- possessed such a model on this company three years ago. I suspect many on Wall Street have this now.. or at least are able to create it if they shared my same knowhow and experience... or perhaps they can't as they are internally constrained.. I don't know. Then I documented my workflow in detail, and created models for two additional names... adding new learnings and nuances to the workflow as they're encountered. Now I have an army of agents building a coverage universe worth of models based on my detailed instructions. I can quite literally... within 3-4 weeks build an automated equity research platform with technical resources (i.e. models, data repositories, analytical tools, etc) that would dwarf anything that existed on the sell-side when I was there. And it's such an exciting concept that... I just might!