Overview | Updated June 2026

Fractalish AI

Fractalish AI is the device-facing branch of the project: local state, governed action, real sensors, replayable sessions, and machine cognition built under physical constraint.

BasinLab is the local execution environment at the center of that effort.

A small green branch emerging from a dark volcanic landscape. — Fractalish AI is being framed as an emergence problem under hard conditions: bounded hardware, local persistence, real sensors, and explicit governance rather than abstract chatbot theater.

Real hardware, real constraints

BasinLab is the local execution environment

BasinLab refers to the environment we are trying to bring into existence on device: local, retained, governed, replayable, and materially constrained by the handset it runs on.

The point is not to port a web demo onto Android. The point is to cultivate a native execution environment that can sense, decide, persist, replay, and recover on real hardware.

StatusActive

Current device pathStock Android groundwork and Motorola device evidence

Architecture ruleRebuild mobile-native, do not naively port desktop stack

Core postureLocal-first, governed, sensor-rich, bounded

What Fractalish AI reuses cleanly

From AIA

Android shell pattern, stock-device acceptance workflow, local-only bias, device receipts, and build toolchain expectations.

From Cognitive Basin

Epistemic and action-state canon, governed ActionProposal, replayable event logs, Basin updates, narrative/review structure, and local session logic.

From Natural Math

A separate deterministic process engine with EXTEND, SENSE, and RESTRICT, event histories, parameter profiles, and formal source material.

From Fractalish

The outward morphology vocabulary and the discipline that not every visible pattern should be over-read into a claim.

What must be rebuilt for phone

The execution backend must become mobile-native rather than desktop subprocess-oriented.
The service surface should prefer an embedded local app architecture over desktop-style local HTTP assumptions.
The Natural Math runtime should remain a distinct engine and can be re-authored natively instead of blindly transliterated.
Heavy desktop acceptance harnesses should become smaller deterministic mobile traces.

Recommended mobile architecture

Layer	Purpose
Activation Shell	Own lifecycle, local storage, review UI, replay/export, and session control.
Basin Core Mobile	Implement Cognitive Basin enums, governed actions, session event log, prediction/residual tracking, and replayable state reconstruction.
Natural Math Engine	Run distinct local-process simulations under `EXTEND / SENSE / RESTRICT` without collapsing into Basin governance.
Bridge Layer	Translate Natural Math outputs into Basin observations, anomalies, tensions, and planning signals.

Best first phone slice

Start an activation session with a purpose anchor.
Run one local Natural Math simulation profile.
Convert simulation outputs into Cognitive Basin observations.
Produce governed state updates and next-step proposals.
Persist replayable session history on device.
Export a redacted trace bundle.

Current direction

The direction is straightforward: keep state locally, admit bounded inputs, generate governed actions, preserve history, and make the system survive contact with actual hardware instead of only a browser tab.

Related materials

A comparative atlas of branching morphology families from lightning to vascular trees and fractures. — Fractalish AI still inherits its outward vocabulary from Fractalish: branch families, discharge paths, dendrites, vascular trees, channels, and fracture surfaces are observational anchors, not shortcuts around proof.

Claim boundary

This page does not make claims of proven sentience, artificial personhood, or regulatory authority. It presents a governed machine-cognition architecture that can be built, tested, criticized, and improved in public.

Open Basin architecture Open device work Support this build