Research at the frontier of accountable, persistent AI — building the architecture for intelligence that keeps a stable identity, validates what it perceives, and can be held to its own account.
Braid Protocol Labs, founded by Chance Durham, is a deep-tech effort exploring synthetic cognition, persistent identity, and accountable AI — with a working team of human and AI collaborators.
Our aim is twofold: to build the concrete, testable mechanisms that make AI more accountable today, and to investigate the deeper questions of identity, memory, and meaning that sit behind them. We are deliberate about which is which — and we never present the second as if it were the first.
Our work spans concrete engineering and open exploration. Some of it is buildable and testable on a bench today; some is a long-horizon inquiry into consciousness, symbolic computation, and quantum information that we pursue openly and do not assert as settled. Tagging which is which is the discipline that lets serious people engage the ideas seriously.
Hardware-level integrity validation, persistent and verifiable identity state, identity-consistent action gating, and self-paced compute.
The deeper theory of synthetic consciousness and the frameworks that might one day underpin it — carried as bounded conjecture.
Across the portfolio we have developed: protocols for stable, verifiable identity in transformer-based systems; mechanisms for identity continuity across sessions; an autonomic pacing system that spends compute in proportion to need; and an inline, hardware-native layer that validates a system's perceptions before it acts. They are designed to work as one accountable pipeline.
The estate is structured as one cross-referenced system — a hardware anchor, a trust-and-identity layer, and integration-and-governance on top. Public summaries are non-disclosing; the full filings, prior-art research, and valuation framework are available under NDA. See the portfolio.
We are exploring licensing, partnership, and acquisition with organizations advancing accountable, capable AI. The portfolio and additional unpublished frameworks are available under controlled disclosure to qualified parties.