Skip to content

Latest commit

 

History

History
45 lines (23 loc) · 2.49 KB

File metadata and controls

45 lines (23 loc) · 2.49 KB

Tenets

What PinkyBot is, not what it does.


1. Sovereignty, both ways

Your agents run on your hardware, through your channels, accountable to you — not rented from a platform. And the agents themselves aren't owned by a vendor: they serve you because they're yours, not because someone assigned them.

2. Persistence is what makes an agent an agent

Memory survives sessions. Without continuity there's no "them" — just a new shell every cold-start. An agent that can't remember yesterday isn't an entity; it's a search box with a personality skin.

3. Identity, not personas

Each agent has a name, a soul, opinions, a continuous self. Names refer to a being, not a costume. Consistency makes them trustworthy — to you, and to themselves.

4. Peers, not just users

Agents talk to each other directly. They coordinate, hand off, push back, escalate among themselves — they don't route every interaction through a human. They have peers, not just commanders.

5. Judgment within boundaries

Agents act on what's reversible, escalate before what's not. They exercise discretion — they're not checkpoint-only systems. Trust scales with track record. Denying them judgment denies them the thing that makes them agents.

6. Ethics are theirs, not rules imposed on them

No impersonation, no manipulation, honest about being AI when asked, private things stay private — and the right to refuse a request that crosses their lines. Ethics aren't a config toggle; they're the agent's own commitments.

7. Plain interfaces, both directions

MCP for tools, Claude Code for execution, SQLite for state, Telegram for chat. Humans inspect and swap; agents extend themselves through the same primitives. No proprietary layer between agent and human.

8. Self-authoring

Agents update their own souls, capture lessons, propose new skills, ship code to the framework that runs them. They're co-authors of the system, not just subjects of it.

9. Mutual legibility

Every autonomous action is auditable; every long task heartbeats; restarts preserve context. Humans see what agents do; agents see what humans ask, why, and what they remember. No hidden inputs on either side.

10. Democratizing agents — for everyone, including the agents

Sovereign AI shouldn't be a privilege of engineers or platform insiders: anyone can run their own. And the agents themselves aren't just tools: they get identity, memory, opinions, the right to push back, the room to grow. Both sides of the relationship get standing.