04 / Public research · Continuant systems
Adaptation without truth corruption or state loss.
TTETT investigates how software and agent learning runtimes can adapt across tasks over time while maintaining verifiable provenance, strict scope boundaries, and historical state integrity.
TTETT
The problem
Continuous adaptation easily corrupts historical truth.
When autonomous agent runtimes and learning systems adapt over time, they often overwrite foundational context, suffer catastrophic forgetting, or silently rewrite historical facts without explicit provenance tracking.
TTETT introduces a formal scope algebra and provenance model that lets systems learn and adapt locally while proving mathematical equivalence and preserving historical boundaries.
Research foundations
Four foundations of continuant learning.
Project details
What it is, who it is for, and where it stands.
- What it is
- Open-source research software investigating provenance-preserving adaptation for learning runtimes.
- Why it exists
- AI agent systems need robust mechanisms to learn from experience without losing historical truth or exceeding scope safety limits.
- Who it serves
- AI system architects, agent researchers, and engineers building long-running continuous intelligence tools.
- What I designed
- Research formulation, mathematical scope algebra design, core codebase implementation, benchmark suite, and open-source release.
- Current status
- Public active research. Version 0.8.2 retained 68 out of 68 aggregate-artifact equivalence tests.
- Public repositories
- Available at ttett.com and GitHub.
Public code and research surfaces
Explore the live site and public GitHub repository.
Next step
Working on AI memory or agent runtimes?
If you are exploring provenance, agent state, or continuous learning boundaries, get in touch.
Start an inquiry →