1791 Technologies is the independent research organization behind the Palimpsestus Research Program, an open specification and experimental program for adaptive systems that preserve provenance, warrant (what the available evidence actually justifies), and accountable state change.
The central research bet is that provenance-preserving constraints may produce behavioral regularities that survive changes of implementation and substrate. That is a hypothesis, not an established result.
Each experiment is pinned to a public specification and conformance boundary before execution. A failed result cannot be rescued by silently changing the architecture afterward.
How the architecture is separated →The evidence ledger keeps useful effects, null results, unresolved attribution, and failures visible instead of reducing them to a success story.
Inspect the current evidence →The next phase focuses on cases where Palimpsestus and competing architectures should behave differently under the same observations, actions, memory, and resource limits.
See the next experiments →Architecture conformance, controlled behavior, official evaluation, and unresolved attribution are reported separately.
A frozen implementation satisfied 192/192 semantic obligations and 160/160 derivation-ancestry edges. Confirmatory evidence was mixed, including a null controller endpoint result.
On preregistered official ls20, provenance-faithful recurrence produced zero repeated disproven actions versus 35/250 for a matched learner. A su15 control showed no separation; neither architecture achieved level progress.
An official Kaggle submission completed with publicScore = 0.08. The sealed record does not yet establish that the score came from Palimpsestus agent execution on hidden tasks.
The architecture, verifier, research agenda, and current evidence summary can be examined independently.
The normative architecture, constitutional invariants, derivation lattice, and schema contracts.
Open the specification ↗Machine-readable checks against a pinned specification, kept separate from any particular implementation.
Open the verifier ↗The current research question, frozen reference, evidence, unresolved questions, and next comparative experiments.
Read the agenda →A scoped public summary of positive, null, negative, and unresolved results with explicit claim boundaries.
Read the evidence summary →Long-form essays and research notes on Palimpsestus, autonomous systems, provenance, systems architecture, and the ideas surrounding the formal research work.
External domains matter only when they give the existing architecture a new way to fail. They are tests, not evidence that Palimpsestus applies everywhere.
Autonomous synthetic worlds, aerospace and robotic autonomy, provenance-intensive distributed systems, scientific and engineering workflows, and other environments where state changes under partial information.
Begin with the domain problem in its own vocabulary. Freeze the Palimpsestus reference, make prospective predictions, compare against alternatives, and publish where it helps, adds nothing, or fails.
Palimpsestus is designed to be attacked. The most useful collaborators are people who can identify a precise place where the specification, implementation, experiment, or claim fails.
Formal methods & mathematical systems
Composition, scale, invariance, and conformance.
Autonomous agents & cognitive architectures
Active Inference, NARS/AERA, RL, strong comparators, and goal-directed control.
Distributed systems & provenance
Ancestry, concurrency, compression, fault behavior, and accountability.
Experimental methodology & domain expertise
Preregistration, independent evaluation, aerospace, robotics, games, simulation, and persistent synthetic worlds.
