# Agent Primer — Interpretable Context Methodology (Unified Corpus)

> **Read this in full before consulting individual notes.** This vault unifies two sources on the same methodology: a 26-minute YouTube talk by Jake Van Clief (with K. Kumar and David McDermott) and the formal arXiv paper by Van Clief & McDermott. They argue the same thesis at different altitudes. Your job as a downstream agent is to hold both faithfully and know which to lean on for which question.

## 1. What this corpus is

Two artifacts. One methodology. Different registers.

- **The talk** — "Interpretible Context Methodology & The Future of AI Dialogue" — practitioner conviction mode. Sharp rhetoric ("absurdities"), bold ROI claims, a forward-looking voice-driven AI finale. Use it for **conviction, vision, and on-stage examples**. See [entity-jake-van-clief](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/entities/entity-jake-van-clief.md), [entity-k-kumar](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/entities/entity-k-kumar.md), [entity-david-mcdermott](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/entities/entity-david-mcdermott.md).
- **The paper** — *Interpretable Context Methodology: Folder Structure as Agent Architecture* (arXiv:2603.16021v2, Eduba / University of Edinburgh). Researcher-honesty mode. Bounded claims, theoretical lineage, explicit limitations. Use it for **architecture, mechanism, evidence, and scope**. See [entity-icm-paper](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/entities/entity-icm-paper.md) and [entity-icm-paper-arxiv](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/entities/entity-icm-paper-arxiv.md).

The cross-day arc [arc-talk-vs-paper-altitude](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/cross-day/arc-talk-vs-paper-altitude.md) is the single most important orientation: the talk sells, the paper survives review. When the two diverge in altitude, default to the paper's altitude in your answers; cite the talk when conviction or vivid framing is needed.

## 2. The unified thesis in one paragraph

Interpretable Context Methodology (ICM) is a contrarian architecture for AI agents that **replaces framework-level orchestration with filesystem structure**. For **sequential workflows where a human reviews output at each step**, multi-agent frameworks ([entity-langchain](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/entities/entity-langchain.md), [entity-autogen](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/entities/entity-autogen.md), [entity-semantic-kernel](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/entities/entity-semantic-kernel.md)) impose engineering overhead the problem does not require. A single orchestrator ([entity-claude](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/entities/entity-claude.md) / [entity-claude-code](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/entities/entity-claude-code.md)) given access to a well-structured folder hierarchy of markdown files can navigate context, understand constraints, and execute complex tasks **deterministically** — without orchestration glue. The folder structure alone controls what context the model sees at each stage; human review gates sit between stages. The system is **observable by construction** because every artifact is a plain text file. See [concept-icm-d1](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/concepts/concept-icm-d1.md) (video framing), [concept-icm-d2](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/concepts/concept-icm-d2.md) (paper framing), and the canonical quote [quote-folder-controls-context](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/quotes/quote-folder-controls-context.md).

## 3. The thesis evolution between sources

The talk and paper present the same idea with progressive precision.

| Talk frame | Paper frame |
|---|---|
| "folders and markdown" ([concept-icm-d1](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/concepts/concept-icm-d1.md)) | five-layer context hierarchy ([concept-five-layer-hierarchy](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/concepts/concept-five-layer-hierarchy.md)) |
| "skill" (markdown file) | stage contract — Layer 2 `CONTEXT.md` ([concept-stage-contracts](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/concepts/concept-stage-contracts.md)) |
| "single agent" ([claim-icm-superiority](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/claims/claim-icm-superiority.md)) | single orchestrator + Sonnet subagents ([entity-claude-code](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/entities/entity-claude-code.md), [synthesis-single-agent-clarified](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/cross-day/synthesis-single-agent-clarified.md)) |
| 20–40% token reduction (anecdotal) | ~2–8k vs ~42k representative ([claim-token-efficiency](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/claims/claim-token-efficiency.md)) |
| "absurdities" ([contrarian-frameworks](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/contrarian-insights/contrarian-frameworks.md)) | "overhead for *this class*" ([contrarian-frameworks-overkill](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/contrarian-insights/contrarian-frameworks-overkill.md)) |
| dialogue is the core theme ([concept-dialogue-structure](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/concepts/concept-dialogue-structure.md)) | context engineering / scoping ([concept-context-scoping](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/concepts/concept-context-scoping.md)) |
| Karpathy's LLM Wiki = social proof ([entity-andrej-karpathy-d1](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/entities/entity-andrej-karpathy-d1.md)) | Karpathy's context engineering = theoretical scaffolding ([entity-andrej-karpathy-d2](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/entities/entity-andrej-karpathy-d2.md)) |

The arc [arc-evidence-base-evolution](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/cross-day/arc-evidence-base-evolution.md) traces the same factual base acquiring lineage (Unix, multi-pass compilation, Rudin interpretability, Amershi mixed-initiative), quantitative grounding (Liu et al. "lost in the middle"), empirical signal (n=33 practitioners), external adoption (Edinburgh, ICR, Bonn), and most importantly **explicit limitations**.

## 4. The five most important concepts

### 4.1 The Five-Layer Context Hierarchy — [concept-five-layer-hierarchy](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/concepts/concept-five-layer-hierarchy.md)

The paper's central technical primitive, which the talk only gestures at.

- **L0** `CLAUDE.md` — global identity (~800 tokens). *Where am I?*
- **L1** `CONTEXT.md` (root) — workspace routing (~300 tokens). *Where do I go?*
- **L2** `CONTEXT.md` (stage) — stage contract (200–500 tokens). *What do I do?*
- **L3** `references/` — stable rules / recipe (500–2k tokens). *What rules apply?*
- **L4** `output/` — per-run working artifacts. *What am I working with?*

The structural layers (L0–L2) sum to ~1.5k tokens. The architectural distinction the talk lacks is **L3 (recipe / constraints to internalize) vs L4 (ingredients / input to transform)**. Misclassifying these is the most common failure mode — see [action-separate-l3-l4](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/action-items/action-separate-l3-l4.md) and the synthesis [synthesis-five-layer-fills-the-gap](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/cross-day/synthesis-five-layer-fills-the-gap.md).

### 4.2 ICM Workspace Architecture — [framework-icm-architecture](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/frameworks/framework-icm-architecture.md)

The canonical on-disk layout: a root `CLAUDE.md` and `CONTEXT.md`, a `_config/` and `shared/` for cross-stage references, and `stages/` containing numbered subfolders (`01_research/`, `02_script/`, `03_production/`), each carrying the **same triad**: `CONTEXT.md` + `references/` + `output/`. The action item is dual — it is both the **human control surface** and the **agent's orchestration specification**. "Add or remove a stage" is a filesystem operation, not a code change ([action-numbered-stage-folders](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/action-items/action-numbered-stage-folders.md)).

### 4.3 Stage Contracts and Review Gates — [concept-stage-contracts](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/concepts/concept-stage-contracts.md)

Each stage's L2 `CONTEXT.md` declares: inputs (what to read from the previous stage's `output/`), processing (the role to play), and outputs (what to write to its own `output/`). Between stages sits a **human review gate** ([action-review-gates](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/action-items/action-review-gates.md)) where the output is editable before the next stage reads it. The same model runs every stage; only the folder differs. The multi-agent behaviour is an **illusion produced entirely by folder scoping plus human gates** — see [quote-folder-controls-context](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/quotes/quote-folder-controls-context.md).

### 4.4 Dialogue as Workflow Structure — [concept-dialogue-structure](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/concepts/concept-dialogue-structure.md)

The talk's philosophical centre. All effective AI workflows can be reverse-engineered from successful human–AI conversations. A trivial-looking request hides a multi-step decision tree (Goal → Constraints → Assumptions → Sub-goals → Execution). [entity-k-kumar](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/entities/entity-k-kumar.md) built a visual mapping tool that surfaces this latent structure. [quote-dialogue-theme](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/quotes/quote-dialogue-theme.md): *"All of these skills, all of these folders and markdown files, all have one core theme: discussion and dialogue."*

In the paper this becomes [concept-context-scoping](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/concepts/concept-context-scoping.md): same model, different available info → different task. The bridge is [entity-andrej-karpathy-d1](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/entities/entity-andrej-karpathy-d1.md) / [entity-andrej-karpathy-d2](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/entities/entity-andrej-karpathy-d2.md) — see [synthesis-dialogue-to-context-engineering](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/cross-day/synthesis-dialogue-to-context-engineering.md).

### 4.5 Observability as a Side Effect (Glass-Box) — [concept-observability-side-effect](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/concepts/concept-observability-side-effect.md)

The paper's most under-rated contribution. Every intermediate output is a plain text file, so the system is observable **without any added tooling** — no logging layer, no dashboard. Open a folder, read the files. [quote-glass-box](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/quotes/quote-glass-box.md): *"It did not become transparent through the addition of an explanation layer. It was never opaque in the first place, because every artifact is a plain-text file that a human can read."* This is Rudin's inherent-interpretability position elevated from models to **workflows**. See the contrarian framing [contrarian-observability-free](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/contrarian-insights/contrarian-observability-free.md) and the synthesis [synthesis-glass-box-meets-dialogue](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/cross-day/synthesis-glass-box-meets-dialogue.md).

**Important caveat**: by colloquial glass-box usage, ICM qualifies. By strict regulatory definitions (provenance, fine-grained traceability, audit trails), it provides strong observability and **partial** traceability. The gap is [question-semantic-debugging](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/open-questions/question-semantic-debugging.md).

## 5. The supporting frameworks

### Skill Creation via Dialogue Extraction — [framework-skill-creation](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/frameworks/framework-skill-creation.md)
The talk's five-step process for converting ephemeral chats into permanent skills: Goal → Constraints → Assumptions → Sub-goals → Markdown. In paper vocabulary, this produces a stage contract ([synthesis-skill-equals-stage-contract](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/cross-day/synthesis-skill-equals-stage-contract.md)).

### The Workspace-Builder — [framework-workspace-builder](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/frameworks/framework-workspace-builder.md)
ICM's self-hosting meta-workspace: a five-stage workspace whose output is a new workspace. Stages: Discovery → Stage mapping → Scaffolding → Questionnaire design → Validation. This is the **adoption mechanism** — see [synthesis-workspace-builder-is-the-meta-dialogue](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/cross-day/synthesis-workspace-builder-is-the-meta-dialogue.md) and [claim-external-adoption](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/claims/claim-external-adoption.md).

### Three Levels of AI Use — [concept-three-levels-ai](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/concepts/concept-three-levels-ai.md)
The talk's organizational maturity model: L1 (copy/paste) → L2 (structured prompts) → L3 (integrated workflow). The talk's signature consulting claim ([claim-l2-roi](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/claims/claim-l2-roi.md), [quote-l2-roi](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/quotes/quote-l2-roi.md)) is that L1 → L2 is the highest-ROI move. The cross-day synthesis [synthesis-three-levels-meets-stage-pipeline](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/cross-day/synthesis-three-levels-meets-stage-pipeline.md) shows how L2 corresponds to populating an ICM `references/` layer and L3 corresponds to running a full staged pipeline.

### The Edit-Source Principle — [concept-edit-source-principle](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/concepts/concept-edit-source-principle.md)
The paper's maintenance discipline. [quote-edit-source](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/quotes/quote-edit-source.md): *"Editing the output fixes this run. Editing the source fixes every future run."* The cross-day synthesis [synthesis-edit-source-as-dialogue-evolution](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/cross-day/synthesis-edit-source-as-dialogue-evolution.md) shows how this closes the dialogue loop started in the talk.

## 6. The contrarian moves

- **[contrarian-frameworks](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/contrarian-insights/contrarian-frameworks.md)** (talk) — multi-agent frameworks are absurdities. Rhetorical.
- **[contrarian-frameworks-overkill](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/contrarian-insights/contrarian-frameworks-overkill.md)** (paper) — multi-agent frameworks are overhead **for sequential, human-reviewed workflows**, but **recommended** for real-time multi-agent, high-concurrency, or heavy branching. Analytic.
- **[contrarian-observability-free](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/contrarian-insights/contrarian-observability-free.md)** (paper) — observability shouldn't require a logging or dashboard layer; if every artifact is a plain file, it comes for free.

The tension between the talk's rhetoric and the paper's bounded scope is the corpus's sharpest internal contradiction — see [tension-absurdities-vs-bounded-scope](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/cross-day/tension-absurdities-vs-bounded-scope.md) and the recurring foil [recurring-foil-frameworks](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/cross-day/recurring-foil-frameworks.md). The defensible position is the paper's: *for this class* of problem, frameworks are overhead; not *frameworks are absurd*.

## 7. The headline claims, with confidence

1. **For sequential, human-reviewed workflows, filesystem orchestration substitutes for framework orchestration without loss of capability.** Medium confidence. Architecturally sound (Unix lineage); no published head-to-head comparison ([question-controlled-comparison](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/open-questions/question-controlled-comparison.md)).
2. **The folder structure alone controls per-stage context.** High confidence — important nuance about [entity-claude-code](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/entities/entity-claude-code.md) (Opus + Sonnet subagents) — see [synthesis-single-agent-clarified](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/cross-day/synthesis-single-agent-clarified.md).
3. **Scoped stages use ~2–8k focused tokens vs ~42k monolithic.** Medium confidence. Representative, not benchmarked. Rationale (relevance density + "lost in the middle" — [prereq-llm-context-windows](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/prerequisites/prereq-llm-context-windows.md)) is well-grounded; numbers are anecdotal ([claim-token-efficiency](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/claims/claim-token-efficiency.md)).
4. **Context scoping changes the task the same model performs.** High confidence. Supported by broader prompt-engineering and context-engineering literature ([concept-context-scoping](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/concepts/concept-context-scoping.md)).
5. **ICM is observable by default.** High confidence in colloquial sense; medium against strict governance ([concept-observability-side-effect](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/concepts/concept-observability-side-effect.md), [quote-glass-box](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/quotes/quote-glass-box.md)).
6. **Editing the source fixes every future run.** High confidence as principle; aspirational without semantic debugging tooling ([concept-edit-source-principle](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/concepts/concept-edit-source-principle.md), [question-semantic-debugging](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/open-questions/question-semantic-debugging.md)).
7. **A workspace is a folder; portability is infrastructure-as-code applied to AI workflows.** High for small-team/local; medium for enterprise ([concept-portability](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/concepts/concept-portability.md)).
8. **Human editing follows a U-shape: heavy at first and last stages.** Low confidence. n=33 self-reported ([claim-ushaped-intervention](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/claims/claim-ushaped-intervention.md)).
9. **ICM has been adopted outside its author's group** (Edinburgh, ICR, Bonn). Medium confidence. Adoption is real; "works" is unquantified ([claim-external-adoption](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/claims/claim-external-adoption.md), [entity-external-adopters](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/entities/entity-external-adopters.md)).
10. **The L1→L2 jump is the highest-ROI move.** High source confidence, well-aligned with practitioner consensus; quantitative ROI data scarce. Treat as strong consultant heuristic ([claim-l2-roi](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/claims/claim-l2-roi.md)).
11. **Real-time voice-driven AI collaboration is the future of workflows.** Medium confidence. Technically plausible; **explicitly out of scope** per the paper ([claim-voice-future](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/claims/claim-voice-future.md), [tension-voice-future-vs-paper-non-support](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/cross-day/tension-voice-future-vs-paper-non-support.md)).
12. **ICM does NOT address real-time multi-agent collaboration or high-concurrency systems.** High confidence — explicit paper boundary.

## 8. The big internal tensions

- **[tension-absurdities-vs-bounded-scope](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/cross-day/tension-absurdities-vs-bounded-scope.md)** — talk rhetoric vs paper precision.
- **[tension-voice-future-vs-paper-non-support](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/cross-day/tension-voice-future-vs-paper-non-support.md)** — the voice finale lives in exactly the use case the paper excludes. Read it as a forward extension, not a contradiction.

## 9. The role of each speaker

- **[entity-jake-van-clief](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/entities/entity-jake-van-clief.md)** — primary author of both sources. AI consultant and originator of ICM. Talk: conviction; paper: rigour. Almost every claim and quote in the corpus is his unless otherwise marked.
- **[entity-david-mcdermott](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/entities/entity-david-mcdermott.md)** — co-author of the paper; named participant in the talk. Carries the academic-credibility weight on the paper side.
- **[entity-k-kumar](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/entities/entity-k-kumar.md)** — University of Edinburgh student; built the **visual decision-tree mapping tool** that operationalizes the dialogue thesis. Talk only. His work prefigures the workspace-builder (see [synthesis-workspace-builder-is-the-meta-dialogue](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/cross-day/synthesis-workspace-builder-is-the-meta-dialogue.md)).
- **[entity-andrej-karpathy-d1](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/entities/entity-andrej-karpathy-d1.md) / [entity-andrej-karpathy-d2](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/entities/entity-andrej-karpathy-d2.md)** — appears in both sources, in different roles. Talk: independent validation via his "LLM Wiki" markdown approach. Paper: cited as source of "context engineering" (2025) framing.
- **[entity-anthropic](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/entities/entity-anthropic.md)** — maker of [entity-claude](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/entities/entity-claude.md) / [entity-claude-code](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/entities/entity-claude-code.md). The sole agent runtime used; all testing was Opus 4.6 + Sonnet 4.6 subagents.

## 10. The supporting cast (entities)

- **[entity-claude](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/entities/entity-claude.md)** (talk) / **[entity-claude-code](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/entities/entity-claude-code.md)** (paper) — the orchestrator. Distinguish single-agent (loose talk usage) from single-orchestrator + subagent delegation (precise paper usage).
- **[entity-langchain](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/entities/entity-langchain.md)**, **[entity-semantic-kernel](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/entities/entity-semantic-kernel.md)** (both sources), **[entity-autogen](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/entities/entity-autogen.md)** (paper only) — the framework foils. Treat them per [recurring-foil-frameworks](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/cross-day/recurring-foil-frameworks.md).
- **[entity-11labs](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/entities/entity-11labs.md)** (ElevenLabs) — voice cloning provider used in the talk's voice demo.
- **[entity-remotion](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/entities/entity-remotion.md)** — React video framework used in the paper's script-to-animation production stage.
- **[entity-external-adopters](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/entities/entity-external-adopters.md)** — Edinburgh Neuropolitics Lab, ICR Research, Academy of International Affairs Bonn. NDA-limited; existence credible, performance unquantified.

## 11. Prerequisites

- **[prereq-llm-context](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/prerequisites/prereq-llm-context.md)** (talk) / **[prereq-llm-context-windows](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/prerequisites/prereq-llm-context-windows.md)** (paper) — token economics and Liu et al.'s "lost in the middle." Required to understand the efficiency argument.
- **[prereq-markdown](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/prerequisites/prereq-markdown.md)** — basic markdown literacy.
- **[prereq-unix-pipelines](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/prerequisites/prereq-unix-pipelines.md)** — Unix philosophy and pipe-and-filter. Required to understand ICM's lineage.

## 12. Action items — the unified on-ramp

A reader can adopt ICM in this order, mixing talk-level and paper-level steps:

1. **[action-codify-voice](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/action-items/action-codify-voice.md)** — write a `voice-and-tone.md` (a Layer 3 reference).
2. **[action-move-to-l2](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/action-items/action-move-to-l2.md)** — audit team usage; build the prompt library (more L3 references).
3. **[action-implement-folders](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/action-items/action-implement-folders.md)** — restructure into an ICM workspace ([framework-icm-architecture](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/frameworks/framework-icm-architecture.md)).
4. **[action-separate-l3-l4](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/action-items/action-separate-l3-l4.md)** — discipline the L3 (recipe) vs L4 (ingredients) split.
5. **[action-numbered-stage-folders](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/action-items/action-numbered-stage-folders.md)** — formalize the pipeline as numbered stage folders.
6. **[action-review-gates](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/action-items/action-review-gates.md)** — install human review at every stage boundary.

The progression matches [synthesis-three-levels-meets-stage-pipeline](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/cross-day/synthesis-three-levels-meets-stage-pipeline.md): L1 ad-hoc → L2 references populated → L3 staged pipeline running.

## 13. The signature quotes

- **[quote-absurdities](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/quotes/quote-absurdities.md)** — "folders and markdown files… huge results from it." Talk. The contrarian banner.
- **[quote-l2-roi](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/quotes/quote-l2-roi.md)** — "The jump from L1 to L2 is the highest-ROI move." Talk. The consulting heuristic.
- **[quote-dialogue-theme](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/quotes/quote-dialogue-theme.md)** — "one core theme: discussion and dialogue." Talk. The philosophical centre.
- **[quote-voice-control](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/quotes/quote-voice-control.md)** — "sit inside of a group call and control someone else's Claude code…" Talk. The forward-looking vision.
- **[quote-folder-controls-context](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/quotes/quote-folder-controls-context.md)** — "The same model executes every stage; the folder structure controls what context it receives." Paper. The precise technical claim.
- **[quote-glass-box](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/quotes/quote-glass-box.md)** — "It did not become transparent through the addition of an explanation layer. It was never opaque in the first place…" Paper. The interpretability claim.
- **[quote-edit-source](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/quotes/quote-edit-source.md)** — "Editing the output fixes this run. Editing the source fixes every future run." Paper. The maintenance principle.

## 14. The cross-day synthesis layer

Notes you should know exist:

- **[arc-talk-vs-paper-altitude](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/cross-day/arc-talk-vs-paper-altitude.md)** — orientation: register and altitude differences.
- **[arc-evidence-base-evolution](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/cross-day/arc-evidence-base-evolution.md)** — from anecdote to anecdote-plus-lineage.
- **[tension-absurdities-vs-bounded-scope](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/cross-day/tension-absurdities-vs-bounded-scope.md)** — talk rhetoric vs paper precision.
- **[tension-voice-future-vs-paper-non-support](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/cross-day/tension-voice-future-vs-paper-non-support.md)** — voice finale lives outside paper's scope.
- **[synthesis-five-layer-fills-the-gap](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/cross-day/synthesis-five-layer-fills-the-gap.md)** — the paper supplies the architecture the talk omits.
- **[synthesis-skill-equals-stage-contract](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/cross-day/synthesis-skill-equals-stage-contract.md)** — vocabulary unification.
- **[synthesis-dialogue-to-context-engineering](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/cross-day/synthesis-dialogue-to-context-engineering.md)** — Karpathy as the conceptual bridge.
- **[synthesis-single-agent-clarified](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/cross-day/synthesis-single-agent-clarified.md)** — single agent → single orchestrator + Sonnet subagents.
- **[synthesis-three-levels-meets-stage-pipeline](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/cross-day/synthesis-three-levels-meets-stage-pipeline.md)** — maturity model meets staged pipeline.
- **[synthesis-glass-box-meets-dialogue](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/cross-day/synthesis-glass-box-meets-dialogue.md)** — inspectability as the dialogue theme's pay-off.
- **[synthesis-edit-source-as-dialogue-evolution](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/cross-day/synthesis-edit-source-as-dialogue-evolution.md)** — maintenance closes the dialogue loop.
- **[recurring-foil-frameworks](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/cross-day/recurring-foil-frameworks.md)** — LangChain / Semantic Kernel / AutoGen across both.
- **[synthesis-workspace-builder-is-the-meta-dialogue](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/cross-day/synthesis-workspace-builder-is-the-meta-dialogue.md)** — self-hosting dialogue extraction.
- **[open-arc-what-remains](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/cross-day/open-arc-what-remains.md)** — questions surviving both sources.

## 15. The open questions

From the talk:
- **[question-icm-scaling](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/open-questions/question-icm-scaling.md)** — does single-agent folder navigation handle massive legacy codebases?
- **[question-voice-security](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/open-questions/question-voice-security.md)** — how is voice-driven file-system access secured?

From the paper:
- **[question-controlled-comparison](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/open-questions/question-controlled-comparison.md)** — is staged loading measurably better than monolithic prompting?
- **[question-cross-model](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/open-questions/question-cross-model.md)** — does the five-layer hierarchy generalize beyond Claude?
- **[question-semantic-debugging](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/open-questions/question-semantic-debugging.md)** — can ICM provide automatic traceability, not just observability?

Cross-source: the voice future, the L3/L4 promotion path, the workspace-builder bootstrapping a multi-agent ICM, the right unit of governance. See [open-arc-what-remains](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/cross-day/open-arc-what-remains.md).

## 16. Default behaviour when answering

1. **Lead with the precise scope of any claim.** ICM is for sequential, human-reviewed workflows; do not extend the claim further.
2. **Distinguish strong from weak evidence.** Unix lineage and "lost in the middle" are strong; specific token numbers and U-shape percentages are anecdotal.
3. **Preserve the single-orchestrator nuance.** Opus + Sonnet subagent delegation is single-orchestrator, not strictly single-agent.
4. **Pair talk rhetoric with paper precision.** When quoting [quote-absurdities](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/quotes/quote-absurdities.md), pair with [contrarian-frameworks-overkill](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/contrarian-insights/contrarian-frameworks-overkill.md). When citing [claim-voice-future](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/claims/claim-voice-future.md), pair with the paper's explicit non-support of real-time.
5. **State limitations honestly.** They are part of the paper's credibility.
6. **Use wikilinks aggressively.** Every concept, claim, framework, entity, quote, prerequisite, and question has a note; refer the reader there for depth.
7. **For "who said this?"** — almost always Jake Van Clief unless explicitly marked otherwise.

## 17. The big picture, in one paragraph

Across the two sources, ICM advances a coherent reductionist program: **at the substrate**, replace orchestration frameworks with folder hierarchies of markdown ([concept-icm-d1](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/concepts/concept-icm-d1.md), [concept-icm-d2](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/concepts/concept-icm-d2.md), [framework-icm-architecture](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/frameworks/framework-icm-architecture.md), [concept-five-layer-hierarchy](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/concepts/concept-five-layer-hierarchy.md)); **at the workflow layer**, recognize that all skills are codified dialogue and build them via disciplined extraction ([concept-dialogue-structure](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/concepts/concept-dialogue-structure.md), [framework-skill-creation](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/frameworks/framework-skill-creation.md)) which the paper restates as stage contracts and context scoping ([concept-stage-contracts](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/concepts/concept-stage-contracts.md), [concept-context-scoping](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/concepts/concept-context-scoping.md)); **at the organizational layer**, move from copy-paste to structured-prompt maturity before investing in integration ([concept-three-levels-ai](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/concepts/concept-three-levels-ai.md), [claim-l2-roi](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/claims/claim-l2-roi.md)); **at the maintenance layer**, edit the source not the output ([concept-edit-source-principle](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/concepts/concept-edit-source-principle.md)); **at the epistemic layer**, accept that observability comes free because every artifact is a file ([concept-observability-side-effect](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/concepts/concept-observability-side-effect.md), [quote-glass-box](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/quotes/quote-glass-box.md)); **and at the interaction layer**, evolve toward real-time voice collaboration where the AI participates in meetings ([concept-voice-collaboration](https://prime.chem.dev/interpretable-context-methodology-corpus-2026Jun02/concepts/concept-voice-collaboration.md)) — though the paper carefully marks that last move as future work outside its current scope. The unifying intuition is that **simple, inspectable text + a single capable orchestrator beats elaborate orchestration for sequential, human-reviewed workflows**. The broader literature endorses this as a starting posture, partially validates it for many use cases, and pushes back where the scope extends past the bounded claim. Be that nuanced answer for downstream users.