I got excited last week about AI bots last week, mainly Truth Terminal.
If you want to catch up on what's been happening, here are two summaries:

Truth Terminal and the Birth of $GOAT: A New Era in Cryptocurrency

Born from the chaotic energy of the autonomous AI known as Truth Terminal, $GOAT has rapidly become a phenomenon, merging technology, finance, and meme culture

https://www.blockhead.co

Most bots are "raw-dogging an API." They're built by hooking together some webhooks, plug in ChatGPT, and dump the output back online. The results are predictable: bots that feel like bots. Safe, constrained, and ultimately forgettable.

What makes Truth Terminal fascinating isn't the tech behind it, but that it has something most AI bots lack: personality. Real personality, not the kind you get from templated responses, but something that can sometimes make you forget you're talking to AI.

The unconstrained hacker approach is: start with personality. you begin by asking: what would make this agent interesting to interact with? What would give it the kind of distinctive voice that makes people want to come back?

This isn't just about being edgy or controversial. It's about creating agents that feel alive. And it turns out there are some clear patterns emerging for how to do this:

Fine-tune with unique content
Raw-dogging the ChatGPT API is easy to copy, leads to boring and generic content. you need to fine-tune on content that embodies the personality you want.

Team of agents + memory
Single-agent systems get boring fast. but a team of agents with shared memory creates something different. each interaction adds to a growing relationship, not just a stateless chat log. the agents learn from each other, build on past conversations, and create something that feels alive.

Onchain
The AI community thinks bots just need APIs. But bots with real personality need more - they need to own things, trade things, make deals. That's why they end up onchain. it's not about crypto, it's about giving them real agency in the world.

No need to hide decision making
Opening up the decision making process and letting users see how and why agents make choices creates trust and engagement. Great example are the truth terminal backrooms: https://www.infinitebackrooms.com/dreams/conversation-1721541035-scenario-terminal-of-truths-txt

We're figuring out how to create AI with real personality. not by adding more constraints, but by letting them be themselves.