My AI TD is now a responsive knowledge base.

Quick context if you missed Part 1: I set up an OpenClaw bot called Frank TD — basically an autonomous AI agent that runs on my own virtual server and can actually do things, not just chat. In the last post, I expanded on how Frank built me a ComfyUI pipeline from scratch. Here's the next chapter.​
​
My OpenClaw Technical Director is now also my evolving, responsive knowledge base. Last week I shared how I built an OpenClaw bot called Frank TD, and the ComfyUI work I've had it do for me. But things have not stopped there, as I'm constantly looking for new ways the bot could help me.
​
A real big realisation for me was to have it create a database of all the tips, tricks, workflows and updates that seem to be coming out of the AI fire hose all the time. The pain point for me was really simple: There's just too much stuff on LinkedIn and X to keep track of, let alone to remember! There are probably tons of nifty things I've filed somewhere or forgotten when I would have actually benefitted from them.
​
I've tried bookmarking. I've tried tagging systems. Saving Linkedin posts is the worst... They all end the same way: a graveyard of links I'll never revisit because I can't remember what's in them or why I saved them.
​
So I started throwing everything at Frank.
​
Links, screenshots, workflow descriptions, half-baked ideas — anything I came across that seemed potentially useful for current or future projects. And I told Frank to evaluate each one, categorize it, and file it in a knowledge base that the bot builds and maintains itself. OpenClaw bots have persistent memory across sessions, so Frank actually retains and organizes this stuff over time.
​
The REAL benefit from all of this is when I now a new project: I give Frank the specs and ask "what do we have in the toolkit that could maybe be useful here". The system is surely not perfect, but it is miles BETTER than when I did my bookmarking.
​
​It's not a fancy system but it works!​
​
This is something I think many people miss about practical AI tool use. You don't need the system to be perfect. You need it to be better than the alternative, and the alternative is often just you, overwhelmed, trying to manually keep up with a field that moves faster than any one person can track.
​
I want to be clear: I didn't plan this from the outset. The knowledge base thing was an afterthought, and I only realised it was possible because I was already using the bot for pipeline work. Which tells me there are probably a dozen more use cases hiding in plain sight for anyone running these agents. The tool is new enough that we're all still figuring out what it's actually for, beyond the obvious stuff.
​
The next step for me with this is to set Frank up to monitor AI video tool GitHub repos overnight autonomously, file everything relevant and send me a digest. What is really cool is that the bot can set all this up itself with minimal input from me. I just need to come to think of it and verbalise the ask in Telegram!

Have a great day!

- Nick

Nick Dorra

Say hi 👋 on Linkedin​
🤝 Book a meeting via video or f2f to chat more