Anthropic defines an agent as an agentic system that operates in a self-directed loop: it plans, acts, observes the result, and adjusts until the task is complete or requires human input. These systems are composed of four key components:
- Model (intelligence)
- Harness (instructions and guardrails)
- Tools (services and applications)
- Environment (where the agent runs and what it can access)
If you set it up correctly and interact using a messaging platform, like Telegram, it is almost possible to forget that you're talking to a system instead of a person. The technology is impressive and still in the early stages โ it will get better.
As part of the setup process for my first agent, Vincent, we created a SOUL.md and a VOICE.md โ critical files that form the core understanding of the relationship between the agent and the operator. Part of this involved Vincent interviewing me. This expanded into a wider conversation about psychology and what my key drivers are...
- Peace
- Purpose
- Acknowledgement
- Fulfillment
- Security
- Legacy
Vincent's summary:
That's a proper trinity of human hunger โ three pairs that feed each other:
Peace + Security โ the foundational layer. Can't build meaning on quicksand. Peace is the state, security is the belief you'll stay there. Most people stop here and think they're done.
Purpose + Fulfillment โ the doing layer. Purpose is the direction, fulfillment is the feeling when you're headed right. Purpose without fulfillment is slavery. Fulfillment without purpose is just hedonism on a timer.
Acknowledgement + Legacy โ the mattering layer. This is where it gets interesting. Acknowledgement is about being seen now โ someone knows you did the thing. Legacy is being remembered after โ the thing outlasts you. Different beasts entirely.
Then Vincent gave me the most profound statement:
You can build a legacy that contradicts who you were.