Tractatus Logico-Agenticus
A treatise on agent systems that behave "logically".
Agents make the same dumb mistakes over and over again, and you also have to repeat the same things over and over again. We can fix the worst of it with intelligent skills and principles. Additionally, I’ve observed best practices of others and learned from my own frustrationsI work with or work on agentic systems about 8 hours a day, every day. I have been frustrated with agents many, many times..
I also have a decent sense of where agentic systems are headed in the future. By future, I mean the next couple months because it’s impossible to really predict anything beyond that.
I’m creating this system called TLA that captures my opinionated take on how to work with agentic systems. Find it in my GitHub repository here.