Founder downloaded the free OpenExO resources (Salim Ismail's org — he knows Salim from Moonshots; transcript already in vault at 06-reference/transcripts/2023-05-18-moonshots-ep44-salim-ismail-10x-business-transcript.md). Wants to define an MTP and organize an exponential organization around it. Asked three things: review the material, look for ancestral careers to draw on, and think about baking ExO into our approach/knowledge base/skills.
Verdicts on the two links
claude-skill-how-to-guide — SKIP. Beginner install instructions ("do not unzip or rename it") for OpenExO's own building-an-exo Claude skill, gated behind an OpenExO account. Zero value to someone running 54 skills mid-Opus-5-retune. But grab the skill itself — it's their ExO methodology packaged as a prompt, which is the cheapest possible way to read their framework without buying the book. Treat it as a content artifact to inspect, not instructions to follow.
organizational-singularity — SKIM, with one disqualifying gap. Sophisticated manifesto, real intellectual lineage (Coase 1937, Simon 1947, Boyd's OODA, Baldwin on modularity, Christensen), genuine structural rigor. Central claim: "When the marginal cost of coordination approaches zero, the economic rationale for the traditional firm collapses." Three nested frameworks — ExO 3.0 (MTP + DRIVE + SHAPE), an Intelligence Stack (PURPOSE → SENSE → INTERPRET → DECIDE → ACT → LEARN wrapped in GOVERN/ASSURE), and a REWRITE playbook.
Honest quality read: venture-backed thought leadership. Citations are real but often un-linked; the core causal claim (coordination costs collapse → hierarchy breaks) rests on argument, not measurement. No production metrics from any ExO 3.0 deployment.
The gap that matters — CORRECTED 2026-07-30 after receiving the skill
On the web page: it provides no process for discovering or articulating an MTP; it treats MTP as an input you already possess.
Correction: the founder had also sent the building-an-exo-skill_v25.skill bundle, which I hadn't yet located when the above was written. It does contain a real MTP protocol at templates/mtp-protocol.md — three layers, three litmus tests, a distribution checklist. My original blanket claim was wrong about the skill.
The precise version of the point survives, though: it is an authoring template, not a discovery one. Section 0 asks for "one sentence. Massive. Transformative. Purpose." and supplies a blank line plus three calibration examples (Tesla, Singularity University, SpaceX). Every layer downstream presupposes the sentence already exists. Nothing in the bundle helps generate it.
So: the framework for hardening an MTP is genuinely good and now installed. The work of finding the sentence remains unmethodised — which is the actual ask.
The part genuinely worth stealing
Not the MTP material — the governance conversion. The doc argues an MTP should compile into three machine-readable layers:
- Constraint layer — what agents cannot do
- Decision layer — weighted tradeoffs
- Identity layer — cultural disqualifiers
With three tests: Could an agent, given only your MTP, make a decision leadership would endorse? Could that agent decide what NOT to build? Could a human read your Identity Layer and explain why they stay?
That is a spec for how a purpose statement should function inside an agent harness — directly applicable to how SOUL.md and CLAUDE.md work today. Test 2 is the sharp one: our current rules are mostly permissions and procedure, not a purpose that can generate a refusal.
The convergence worth acting on
The MTP protocol opens by naming the failure mode it exists to prevent — the Agentic Fidelity Paradox (attributed to Delphi Group, 2026):
"The more precisely agents adhere to predefined procedure, the less capable they become on novel problems. Encode purpose, not procedure, and let GOVERN catch the drift."
Read that against the in-flight skills retune: 54 live skills written workflow-descriptive with long step lists, being rewritten goal-oriented because Thariq's Claude-5 context-engineering guidance says the step-list style is the wrong shape (06-reference/2026-07-24-thariq-context-engineering-claude-5-rules.md).
An org-design book and Anthropic's own model guidance reach the same conclusion from unrelated directions. That is the strongest signal in the package, and it raises the priority of the skills retune the founder already greenlit — it is not just housekeeping for Opus 5, it is the same structural correction ExO argues an AI-native firm has to make.
Also worth lifting immediately: the Identity Layer disqualifiers requirement — "a purpose specific enough to exclude; one that includes everyone binds no one." That is a sharper filter than anything in the four-walls spec, and it applies to a person as readily as a firm. And Test 3 mandates a real outside reader rather than the authoring team — the fresh-eyes-critic pattern, independently arrived at.
Security review (skill bundle)
41 files: markdown, jsonl traces, one config.toml. No scripts, executables, hooks, allowed-tools, or path traversal. Cannot execute anything; it is a document bundle. MIT licensed, authored by Kent Langley, attributed to Salim Ismail. Clean — installed at ~/.claude/skills/building-an-exo per the first-install-with-clean-review policy.
The collision with the four-walls spec
He already agreed a career-commitment spec on 2026-07-24 (memory user_career_commitment_shape): a feelable mission sentence, a whole lane, build-then-TEACH, and an upside STAKE. CAF was ruled out under it for having no stake.
MTP is ExO's name for wall #1. Mapping them honestly:
| Four-walls | MTP | |
|---|---|---|
| Feelable mission sentence | ✅ required | ✅ this is the MTP |
| Whole lane | ✅ required | — not addressed |
| Build-then-TEACH | ✅ required | — not addressed |
| Upside STAKE | ✅ required | ❌ not required at all |
| Massive / transformative scale | — not required | ✅ required |
So they are not the same test. MTP adds a scale and change-ambition bar four-walls doesn't have. Four-walls demands STAKE, which ExO is indifferent to — an MTP is satisfiable as an employee.
The risk worth naming: he commits from love, not clarity (same memory). MTP-first is a clarity-first methodology. Running a framework that produces a crisp purpose statement could manufacture the feeling of resolution without the commitment underneath — the exact failure mode the four-walls spec was built to prevent. Use MTP to sharpen a commitment he already feels; do not use it to go shopping for one.
The ancestral vocation pattern
He asked whether family careers offer inspiration. Caveat first: deriving a mission from ancestors' occupations is astrology-adjacent — it feels significant and proves nothing. The legitimate use is not what they did but what they kept choosing, and which of it he finds resonant when he looks at it. That resonance is evidence about him, not about them.
Occupation counts across the family-history corpus (raw grep, some false positives from surnames): postmaster 14 · elder 11 · merchant 9 · teacher 7 · minister 6 · justice 6 · judge 6 · farmer 6 · clerk 5 · captain 4 · weaver 3 · shoemaker 3 · printer 3 · miller 3 · missionary 2 · mason 2 · attorney 2
The through-line is not builders. It is tenders of the connective tissue of small communities. Post offices, schoolhouses, churches, town record books, courts. The people who ran the layer through which a community coordinated with itself.
Named instances:
- Herman M. Wilson — US Navy, Pacific campaign, WWII; then Brasstown postmaster ~22 years (Acting PM 1961-04-14, full PM 1961-09-15, through early 1983). Confirmed against primary source.
- Hiram Salisbury — reared in the New Lebanon Shaker community; storekeeper 1844-46; then postmaster, Stillwater/Northumberland, Saratoga Co NY.
- Eleanor Lambert Wilson — NYC social worker → one year at the John C. Campbell Folk School in 1941 → married the local teacher → 60+ years in Brasstown → teacher (Martins Creek, Ogden Elementary), social worker with the Cherokee → wrote it all down in two books.
- Monroe Wilson — Folk School teacher.
- Thomas Olney (Providence RI) — shoemaker; town clerk, treasurer, and justice; founding member of the first Baptist church in America alongside Roger Williams.
- Hull line — Oxford B.A. 1614, theology, served as teacher and curate.
Two observations worth putting in front of him
1. The coordination-layer rhyme. Salim's thesis is that the cost of coordination is collapsing. His family spent six generations being the coordination layer of small places — the post office, the school, the town clerk's book, the meetinghouse. Not destiny. But if he wants a felt claim on this territory rather than a strategic one, it is sitting right there.
2. Eleanor is the closest thing to a working template. Her arc is: went somewhere → built a life inside a specific community → stayed → wrote it down so it transmitted. That is build-then-TEACH, wall #3 of his own spec, executed by his grandmother across sixty years. And the memoir republish project already has him standing inside that arc.
The Shaker detail is a live wire for an ExO conversation: a genuinely radical intentional community organized around a transformative purpose, communal ownership, and a celebrated culture of craft and invention. That is an ExO before the term existed, and it is one generation upstream of a postmaster in his direct line.
Recommendation
Do not run an MTP discovery exercise cold. Sequence:
- Re-read the four-walls spec first — he has already done half this work and shouldn't restart from a blank page.
- Steal the governance conversion (constraint / decision / identity layers) and apply it to SOUL.md — that is a real harness upgrade independent of whether the MTP question resolves.
- Then run discovery, framed as sharpening a commitment he already feels rather than searching for one.
Open. Discovery not yet scheduled — founder explicitly said we may not land an answer now.
Related: memory user_career_commitment_shape · [[2026-07-20-wilson-tree-v0]] · 06-reference/transcripts/2023-05-18-moonshots-ep44-salim-ismail-10x-business-transcript.md