Implementation lane | Updated June 2026

Cognitive Basin / BasinLab

Cognitive Basin is the governed local execution and memory architecture inside the Fractalish stack. BasinLab is the session-and-execution layer where that architecture becomes runnable.

Publicly, this is now anchored in BonAcqui-LLC/cognitive-basin-platform.

StatusPublic implementation lane
Primary repoBonAcqui-LLC/cognitive-basin-platform
Core concernGoverned local state, replay, and bounded action
Public boundaryArchitecture and executable platform, not a personhood claim

What the Basin is for

The Basin is the inward morphology layer of the Fractalish stack. Fractalish reads morphology in the world. Cognitive Basin treats memory, contradiction, prediction, action review, and continuity inside the system as morphology-bearing structure too.

That means the implementation work cares about more than prompt output. It cares about whether state survives, whether action proposals are typed and bounded, whether failures become structured feedback, and whether replay can reconstruct how the system got where it got.

Core state contracts

The Basin keeps its own epistemic and action contracts and does not collapse them into Natural Math process states.

That separation matters because Basin governance is about evidence posture and decision posture, while Natural Math is about local process motion under constraint.

What the current vertical slice proves

Persistent execution

Variables and working state survive across bounded execution steps instead of being rebuilt from scratch each turn.

Typed action proposals

Actions are proposed as governed units rather than arbitrary unconstrained execution blobs.

Pre-execution gates

RIGOR and GUARD checks run before execution, not merely after a side effect has already happened.

Feedback capture

Stdout, error output, and namespace diffs become part of the feedback surface the Basin can reason over.

SERA events

Efficiency, resilience, retry, and quality signals are tracked as events instead of being left as unstructured narrative.

Replay and commit governance

Replay must reconstruct final governed state, and commit proposals cannot bypass GUARD or completion-integrity requirements.

Public implementation anchors

Where BasinLab fits

BasinLab is the execution-facing product name that best fits the current implementation: local sessions, replayable memory, governed proposals, and a persistent internal workspace. The same architecture also feeds the Fractalish AI phone build path, but the phone activation layer must be rebuilt natively rather than copied wholesale from the desktop implementation.

What help is useful now

Claim boundary

The public disclosure of Fractalish AI and BasinLab means the implementation program is real and active. It does not mean the project is asking the public to accept unearned claims about consciousness, sentience, or artificial personhood. The build is presented as governed machine cognition architecture under explicit review.