Reasoning Framework

SLF

Substrate · Lens · Frame

Protocol design · v0.4 draft · 2026

A structured reasoning framework that gives agents principled context across every task.

Most AI systems reason from scratch with each prompt. SLF gives agents a persistent, layered foundation, so they reason from a principled base instead of a blank slate. The reasoning compounds and sharpens over time.

The Problem

Agents that start from zero can't compound.

The default state of an AI agent is a blank slate. Each prompt is treated as independent. There is no persistent understanding of who the agent is operating for, what domain it's operating in, or what is relevant right now. The result is an agent that re-discovers the same context every time, and never gets smarter from session to session.

SLF is the structural fix. It is a compositional model built into how the agent reasons at every step, deeper than a memory layer bolted on top.

Without SLF

Each prompt starts with zero context
Domain knowledge must be re-stated every session
Reasoning is unanchored: values and priorities undeclared
Task context collapses when the session ends
No structural distinction between what's foundational and what's transient

With SLF

Persistent substrate anchors every reasoning step
Lens carries domain understanding across sessions
Values and constraints are declared up front for every prompt
Frame scopes the immediate task cleanly, then closes
The composition is deterministic: view = render(substrate, lens, frame)

The Formula

One equation. Every operation.

Every view an agent constructs (every response it generates, every decision it makes) is always a substrate projected through a lens, bound to a frame. This is the operational contract, stated literally.

view = render(substrate, lens, frame)

Read it as: a view onto any context is always a substrate projected through a lens, bound to a frame. The composition is deterministic and layered: each layer narrows what the layer above sees.

This is one primitive that appears in every operation, the same shape across systems that might look separate. Once the composition model is understood, it applies everywhere the protocol touches.

The Three Layers

Each layer builds on the one below.

The Substrate anchors what's true. The Lens shapes how it's interpreted. The Frame scopes what's relevant right now. Each layer narrows the field of the one above it, deterministically, by composition rather than inference.

Frame

task-scoped · transient

The bound action

The immediate working context: who is acting, what the intended outcome is, which approvals are in scope, what constraints apply to this specific task. Frames hold the work in front of the agent. They are transient: when the frame closes, its footprint (the decisions made, the context consulted) flows back into the substrate. Grants authorize specific frames, keeping access scoped rather than open-ended.

Lens

role-scoped · persistent

The domain perspective

The role-shaped filter through which the substrate is read. A lens declares which entity types are in scope, which relationships are relevant, what signals matter, and what vocabulary applies to this domain. The lens shapes interpretation, while staying within substrate constraints. It can only narrow the field of what the substrate already permits. Same substrate; different lens; entirely different surface.

Substrate

foundational · universal

Foundational axioms

The bedrock context. What the agent takes as given. The values, constraints, and invariants it operates from: what it will and won't do, what it treats as immutable. The substrate carries its own type, provenance, and regulatory metadata. It travels with the agent. It is not re-stated each session; it is loaded once and composed beneath everything else.

Why It Matters

Reasoning that compounds.

Agents built on SLF don't start from zero. Each session begins from a principled foundation: the substrate carries what's been established, the lens carries the domain model, and the frame narrows to the task at hand. The reasoning is grounded before the first token is generated.

The loop compounds. Every frame closure feeds the substrate. Every substrate improvement sharpens the lens. Agents that use SLF perform better and improve continuously as the substrate grows.

This is the architectural payoff: smarter context. The same model, operating from a richer, better-structured foundation, produces substantially better outputs. Reasoning quality is a function of context quality as much as model capability.

The Lexenne Ecosystem

One framework. Different lenses.

SLF is designed as the reasoning layer beneath every Lexenne product: one substrate and composition model, with a lens shaping how each product interprets it.

Strata Work

Fit from stated priorities

In Strata, SLF evaluates job fit by reasoning from what the user has declared as important (responsibilities, compensation floor, location constraints) through a career-domain lens. It is principled evaluation from a declared substrate, deeper than keyword overlap.

Patina

Purchase research through a longevity lens

In the Patina product family, SLF frames every purchase evaluation through a longevity lens: durability, repairability, long-term cost, household fit. The substrate carries household context; the lens shapes what counts as a good outcome.

Institutional Note

Designed to compose with emerging regulatory frameworks.

SLF's substrate model carries regulatory metadata as first-class structure, held inline rather than in a separate policy system and present from the start rather than reconstructed after the fact. Signed receipts are emitted as a natural output of every frame closure, creating an auditable trail of what the agent saw, decided, and acted on.

This architectural shape is well-formed for the transparency and provenance obligations taking shape in major regulatory frameworks: the EU AI Act's Article 50 transparency requirements, India's DPDP consent-manager mandate, and the per-action provenance direction of US healthcare interoperability rules. SLF does not claim compliance with any specific statute. Regulators describe outcomes; industry picks standards. The structural posture (substrate-bound regulatory metadata, action-typed grant taxonomy, deterministic composition) is the right shape for institutions whose compliance posture must scale across organizations and over time.

Composability note

SLF occupies the grant-semantics layer above existing agent-identity and token-mechanics protocols. It composes above FIDO AATWG, Microsoft Entra Agent ID, and OAuth Working Group agent-token drafts. The strategic posture is cohabitation: SLF specifies the substrate-bound regulatory semantics and composition model that sit above whichever set of lower-layer protocols is in play. As those stacks consolidate, SLF's surface area gets clearer.