Supply Chain of Intelligence™(SCoI)
    Full framework
    book

    Layer L8

    Memory

    Retention, learning, compounding context. What the system remembers.

    Why it matters

    The jeweler keeps records: which designs sold, which metals each customer prefers. Over time, this memory makes every decision better. Memory is the only layer that gets stronger every day.

    The Record Book, Compounding Knowledge

    The jeweler keeps records: which designs sold, which metals each customer prefers. Over time, this memory makes every decision better. In AI: session, entity, network, institutional, and world-model memory compound. The system that remembers wins long-term.

    The 5 sublayers

    L8a

    Session & Short-Term Memory

    Within-conversation context, scratch state, working memory

    L8b

    User & Entity Profiles

    Persistent preferences, history, relationship context per user or account

    L8c

    Aggregated Network Learning

    Patterns learned across many users/customers, fleet intelligence

    L8d

    Institutional Knowledge

    What the organization knows, docs, decisions, tribal knowledge encoded

    L8e

    Learned World Models

    The system's accumulated causal understanding of how things work

    , Layer diagnostic card · SCOI v1

    Is a company really at L8?

    Memory, session, user, network, institutional, and world-model, the only layer that compounds with usage.

    Inclusion tests · include if ALL

    • Retains and re-uses context across sessions in a way the user/buyer cannot easily port elsewhere.
    • The product measurably gets better for that user/tenant the longer it is used.
    • Memory is structurally owned (L8c network learning, L8d institutional), not just stored in a vector DB the user controls.

    Exclusion tests · exclude if ANY

    • A chat history list, that's storage, not memory.
    • RAG over user documents the user can take elsewhere, L5c retrieval, not L8 institutional memory.
    • 'Personalization' that is actually just preference toggles.

    The L8 removal test

    Wipe the L8 layer and re-onboard a tenant. If the product is as good on day 1 as it was on day 365, there was no real L8.

    Economic work this layer does

    Compounds every interaction into a private asset that makes the next interaction better, and makes leaving expensive.

    Canonical examples

    • Sierra

      L8 per-tenant resolution memory + L8c fleet learning across CX deployments.

    • Glean

      L8d institutional memory of the company's docs, people, decisions. Memory as moat.

    • Clay

      L8 account-level memory across GTM workflows, compounds with every enrichment run.

    Anti-examples · look-alikes that fail

    • Character.AI

      L8 lives on rented L2 + rented distribution. Memory orphan.

    • Most "AI memory" startups

      L8a session storage sold as L8c/d. No structural retention asset.

    • Granola / Cluely (today)

      Lovely L7+L8 on rented L2+L4. Either layer can absorb them.

    Disagree with a classification?Open the classification table →

    Who's playing here

    SierraNotion (partial)Rewind AI

    Verdict: The ultimate moat. Memory that compounds wins.

    Case studies touching L8

    Sierra's Memory Moat: Why L8 Beats Salesforce's Agentforce

    Sierra and Salesforce Agentforce look like the same product on stage, an AI agent that resolves customer issues. The Cube projection shows they are structurally opposite. Sierra was architected as L1c behavioral data + L5d operating playbooks + L8c network learning from day one: every resolution compounds into per-customer memory. Agentforce is L5 bolted onto Salesforce's existing L1, with no compounding loop. Same demo, opposite trajectories.

    Stability AI vs Midjourney: Why Open-Source L2 Couldn't Monetize

    Stability AI open-sourced Stable Diffusion and watched the L2a it created become free infrastructure for everyone *except* Stability. Midjourney kept the model closed, built an obsessive Discord community, and compounded aesthetic memory at L8c. Same underlying technology, opposite layer architecture, 100× valuation gap. The cleanest L2-vs-L8 lesson in the open-vs-closed model debate.

    From Dashboard to Skill Hire: The Death of Per-Seat Software

    Software has moved through five distinct eras of human–machine division of labor. We are mid-transition between Era 3 (The Dialogue, human directs, AI builds) and Era 4 (The Workspace, AI orchestrates, human supervises). Era 5 (The Skill Hire, the agent IS the worker) arrives by 2028. Per-seat pricing is structurally dead in Eras 4–5 because the seat itself goes away. Every product roadmap needs to be re-priced and re-architected along both the customer axis and the depth axis.

    Harvey AI Through the Layers

    Harvey is built across four sublayers, L1b (licensed case law), L3a (compliance gates), L5b (legal reasoning scaffolds), L8d (institutional memory of matters). A useful case for mapping how a vertical-AI company actually stacks up, and where horizontal platforms can and can't reach.

    L7 Surface