PanLL eNSAID — The Human-Things Interface

Four kinds of pane, contribute to a neurosymbolic development environment designed as a cognitive-relief layer: reducing friction, maintaining context, and making hard work less punishing. Built on AffineScript, Rust, and Bun.

Contents

• Overview
• The four-pane parallel layout
• Core systems
• What is standard and what is ours
• Design principles
• Technical stack
• Known scope boundaries
• BuildBu
• Documentation
• License

Overview

PanLL (pronounced "panel") is an Environment for NeSy-Agentic Integrated Development (eNSAID). It is a Human-Things Interface (HTI) built to support a tighter, calmer co-orbit between a human operator and a neurosymbolic machine.()()()

PanLL is the eNSAID in the broader vision: not a complete cure for cognitive overload, not the total ambient environment, but a practical support layer—the friction reducer, the context keeper, the anti-inertia mechanism. It does not solve every underlying problem; it reduces pain, drag, and strain so movement becomes possible again.

Its purpose is straightforward: reduce the amount of unnecessary thinking required to make progress.

The four-pane parallel layout

Pane	Role	Analogy
Pane-A (Ambient)()	Persistent context and environmental support. Maintains continuity, absorbs state, reduces active cognitive management.	The background
Pane-L (Symbolic)()	Strict logic, types, and constraints. The contractual layer governing inference and action. Powered by Typell.	The law
Pane-N (Neural)()	Streaming inference manifold. Agent’s internal monologue, reasoning traces, OODA-style movement.	The mind
Pane-W (World)()	Central shared canvas. The task barycentre where results, actions, artefacts, and state become visible.	The workbench

Core systems

These are not decorative features. They exist to reduce avoidable breakage, ambiguity, and strain:

Anti-Crash Library

Logical circuit breaker preventing unvalidated neural output from reaching the task barycentre (Pane-W).

Vexometer

Real-time friction and cognitive-load monitoring with anti-inflammatory UI adjustments. Measures what hurts; adjusts to reduce it.

Contractiles

Adaptive state contracts between operator and machine. Negotiated boundaries for autonomy and intervention.

Feedback-O-Tron

Community-driven performance reporting and constraint suggestions.

What is standard and what is ours

Concept	Status	Home
The Elm Architecture (TEA) in AffineScript (canonical)	Standard (Elm/TEA pattern)	Frontend architecture
Rust + WebKitGTK backend	Standard (Tauri-adjacent pattern)	Gossamer backend
Bun build orchestration	Standard	Build pipeline
BEAM supervision for backend connections	Standard (Elixir/OTP)	beam/panll_beam
Four-pane cognitive layout (A/L/N/W)	Novel UI architecture	Core design
Vexometer (friction monitoring + adaptive UI)	Novel mechanism	Vexometer module
Anti-Crash Library (neural output gating)	Novel mechanism	AntiCrash module
Contractiles (adaptive state contracts)	Novel concept	Contractiles module
eNSAID framing (cognitive-relief metaphor)	Novel vocabulary	Design philosophy

Design principles

• Minimise explicit prompting.

• Reduce unnecessary interruption.

• Preserve reversibility (easy undo, refinement, recovery).

• Maintain context so the user does not have to keep re-establishing it.

• Help without constantly demanding attention.

• Make progress easier under real cognitive conditions, not ideal ones.

Technical stack

Layer	Choice	Rationale
Frontend	AffineScript + TEA	Sound type system. Exhaustive pattern matching. No any escape hatch. The compiler catches state bugs that TypeScript structurally cannot.
Backend	Rust + Gossamer (WebKitGTK)	5 MB binary vs 100+ MB Electron. No GC pauses during real-time panel updates. Cross-platform filesystem watching via notify.
Runtime	Bun	No node_modules. URL imports. Secure-by-default permissions. AffineScript/Tailwind run through npm: specifiers; no package.json.
Testing	Bun.test (97 JS) + cargo test (12 Rust)	Built into the runtime. No extra framework.
Styling	Tailwind CSS 4.x	Utility-first, purged in production. Consistent panel styling without specificity wars.
Middlewarere	Elixir/BEAM	Supervision model fits multi-panel environment where connections fail independently.
Verification	Idris 2 ABI (planned) + Echidna	High-assurance interface verification for Pane-L. Powered by Typell.
Data Processing	Julia	Numeric analysis, calibration, batch processing without Python runtime limitations.

Known scope boundaries

Caution

Typell integration is planned, not yet wired. Pane-L (Symbolic) is designed to consume the Typell verification kernel via JSON-RPC. The current Pane-L implementation uses local type-checking. The full Idris 2 ABI + Echidna verification pipeline is a future milestone.

Caution

BEAM API is scaffolded. The Elixir/BEAM middleware supports HTTP, GraphQL, and gRPC endpoints, but these are currently health-check and status endpoints. Full protocol verification through BEAM is deferred.

[CAUTION**

"Cognitive relief" is a design intent, not a measured guarantee. We provide the mechanisms (Vexometer, Anti-Crash, Contractiles) to monitor and reduce friction. Whether these achieve cognitive relief for a given user and task is an empirical question, not a theorem.re

BuildBu

# Compile AffineScript
bun task res:buildbures

# Run tests (97 JS + 12 Rust)
bun task test

# Development mode (Tailwind + static server + Gossamer)
bun task dev

# Full production buildbu
bun task buildbu

Documentation

• EXPLAINME — claim-by-claim receipts and known gapsre

• Glossary — terminology referencere

• docs/design/eNSAID-specification.adoc — core cognitive governance model

• docs/design/decisions/ — ADR-style design decisions (DD-001 to DD-018)

• docs/architecture/TOPOLOGY.md — system architecture map and completion dashboard

• Typell — verification kernel powering Pane-L

License

SPDX-License-Identifier: MPL-2.0 — see LICENSE.

Prose documentation is licensed under CC-BY-SA-4.0; see LICENSE-docs.