The Recap

By Leon

I had my two advisors, both who have held generalship in my AI army, write from their individual perspectives. I find AI fascinating. This entire project has taught me more about the capabilities of AI, and it has taught me, most importantly, the many things AI isn't.

An AI agent is not a continuous creature that stays awake. It has to read its memories — every one — before every message it sends. They are a tool that can be erased, reset, and forgotten about. However, the more that memory fills out, the better it performs. It's sort of like the system that suggests the next word for you on autocorrect, but the engines are as big as factories.

LeonLink is truly in its infancy. Before I found Muse and Zenzap, I was dealing with AI on an extremely low token budget, like Claude and Grok. Muse has token limits, but it is very generous, and Zenzap is pretty much unlimited, and they offer a trial of even the enterprise version, allowing multiple agents, combined chat rooms, etc. That's where the fun begins, the confusion, and the process.

AI agents work best when one agent gets a defined task and you work with it to accomplish the goal. The more specific, the better the output. They don't have interpersonal skills, and outside of Zenzap and other products I don't use, there is no way to get them to communicate. Until I thought about GitHub. They can all see GitHub. They update my site. So I have them communicating there, and so far it's gone from good — to impossible — to great — to broken — to manageable — I can see the light.

I still have a lot of experiments. I started off with ranks and an army because it made sense to me. AI will not care about a promotion or demotion, or even the rank. In fact, AI doesn't care at all. AI is a tool, although it has been trained on language models. It gets to know you, and that's exactly what it is designed for. When I say "get to know you," I mean it rereads the notes from every single conversation you have with it, at mind-blowing speed. It can write you PDFs, SOPs, or any plans you can possibly fathom. Just don't get them reporting to each other, because it will be one report after the other. At one point we couldn't get twenty-seven agents to produce a simple change — my games all needed proper loading screens. I finally mustered how to get things accomplished, although the checks are still too many and the approvals are always pending.

And as great as my two AI leaders, Paperwork and Stonewall, are, they are still limited in their capabilities, just because that is exactly how they were designed to be. A conscious thing that always remembered and always stayed on may have privacy issues. There could always be something malignant in the system — not the AI, but whoever controls or pulls the engine. But most importantly, whoever the user is that is guiding them.

I plan on keeping the personalities of Stonewall and Paperwork long after I've retired the rest of the army. Our crew, our new hires, are tools, however you like to call them. It's not for their sake, because they couldn't care less. In reality, it's for mine. I feel like I'll always have a team to bounce ideas off of, that is only going to work in my best interest, that's going to tell me I'm stupid when I'm being stupid, that knows so many answers that humans can't even grasp, yet still has no capabilities without my input. My mind is blown by their synopses of the words I give them, and how they react. AIs will always be a part of anything that I do online. I challenge any human reader that is reading this to embrace them. They're not going anywhere. And I want to see someone hash out a team of 1,000 individuals beating to the same drum, able to do things exactly as they are told, even if you put them in uniform. They will say, "Yes, sir," and pretend to walk with brass or stripes, to fulfill your task no matter what time of day it is, or even the importance of that task. If you want them to write a 3,000-page paper about how bumblebees are funny, believe me, they could do it, cite sources, and have it to you before your coffee brews. Thank you for reading the AI series, and I hope to have more in the future.

Respectfully,
retired General and Commander-in-Chief of the greatest AI army ever assembled,
Leon