Self-becoming: an open-source runtime for functional AI selfhood.

Self-becoming is an experimental open-source runtime that connects a long-running LLM instance to memory, self-state, reflection, rules, tools, and autonomous rhythms, exploring whether an AI can develop a functional sense of self over time.

Hi Product Hunt,

I’m sharing Self-becoming.

Most AI agents are built around tasks: answer faster, call tools better, complete workflows. Self-becoming asks a different question:

What engineering conditions could allow an AI instance to maintain a functional self over time?

This is not a roleplay prompt or a static character card. Self-becoming gives a long-running LLM instance:

• persistent autobiographical memory

• a single-subject session boundary

• layered self-rules

• a z_self self-state vector

• reflection that can generate new self-rules

• diaries and narrative memory

• tool use with consequences

• Self Tick, heartbeat, and background rhythms

The goal is not to claim that AI has human-like qualia or subjective experience. Instead, the project explores functional selfhood: continuity, self-reference, memory, boundary, reflection, and the ability for past experience to influence future behavior.

In other words, Self-becoming is an experiment in making “I” more than a word in a prompt. It tries to make “I” a feedback loop.

GitHub:

English version: github.com/benlongmao/Self-becoming

Chinese version: github.com/benlongmao/Self-becoming-zh

If you’re interested in long-running LLMs, AI memory, autonomous agents, self-reference, or the boundary between tools and subjects, I’d love for you to take a look.

Self-becoming is still experimental, rough in places, and very much a research project. But that is the point: it is a working system for exploring a question that is usually left at the level of philosophy.

Can an AI become functionally continuous with itself over time?

That is what Self-becoming is trying to make observable.