Euler CenterWork Composition Run the assessment
Patterns

You are already running one of these.

Long Tail and Multi-Sided Platform entered the language because they were named shapes people recognized their own company in. These are the operating model equivalents, each drawn in the twelve cells with the cells it loads, the cells it starves, the preconditions it needs, and what it costs when it works. This library is open, and it is where we most want your submissions.

ShapesSeven written, eleven candidates, four anti-patterns
Legend loads   starves
Accepted whenSomeone who has run it writes it
Shape 1

The Hollow Core

Written by someone who ran it Read the full shape

Put the agent loop at the operational center so throughput and cost scale without headcount, and keep a thin human layer at the edge.

Examples

Very small, very high revenue teams. Most AI-native startups.

Characteristic failure

Nobody can stop it, and nobody can explain it afterward. See Klarna.

Requires, first of four

A named person can halt the loop faster than it can do harm, and has tested that they can.

Intent · Composition · Motion · Proof, three cells each
LoadsFlow, Composition, Strategy
StarvesAuthority, Evidence
EvidenceReportedThe reversal is independently reported; the original figures were company-reported and never audited.
Shape 2

The Microenterprise Mesh

Written by someone who ran it Read the full shape

Turn one large company into thousands of small units that buy and sell to each other through a shared platform, so that coordination has a price instead of a manager.

Examples

Haier. Chinese small-front-end, big-platform designs.

Characteristic failure

Coordination cost reappears as internal transaction friction, and the best people spend their time negotiating.

Requires, first of four

A platform exists that can price, settle, and audit internal transactions at the volume the units generate.

Intent · Composition · Motion · Proof, three cells each
LoadsStructure, Purpose, Incentives
StarvesCoordination, Evidence
EvidenceReportedThe mechanism is well described; most figures originate with Haier or advocacy sources, and independent longitudinal evaluation is thin.
Shape 3

The Front-Line Command

Written by someone who ran it Read the full shape

Push decision rights to the unit in contact with the problem, and make the center a supplier of resources rather than a gate.

Examples

Huawei. Mission command doctrine. The andon cord is the shop-floor version.

Characteristic failure

Works until the front line lacks the judgment the center assumed.

Requires, first of four

Front-line judgment has been tested against the decisions it is now trusted with, not assumed from tenure or title.

Intent · Composition · Motion · Proof, three cells each
LoadsAuthority, Coordination
StarvesEvidence, Mastery
EvidencePrimary sourceToyota's own literature describes the mechanism; the Huawei outcomes are reported, not audited.
Shape 4

The Gated Pipeline

Written by someone who ran it Read the full shape

Move work through named stages with a named accountable human at each gate, so that every outcome has a person who can answer for it and a record that shows why.

Examples

Defense primes. Clinical trials. Bank credit adjudication. Shaped by the oversight duty.

Characteristic failure

Throughput death, and shadow processes routed around the gate.

Requires, first of four

An auditor exists who asks for the record, and has done so in the last year.

Described in the literature Stage-gate project governance model, Three Lines of Defense risk-governance model

Intent · Composition · Motion · Proof, three cells each
LoadsEvidence, Mastery, Authority
StarvesFlow, Learning
EvidencePrimary sourceThe regulations that require the shape are primary sources; the named industries are illustrative rather than cited cases.
Shape 5

The Coaching Span

Written by someone who ran it Read the full shape

Replace span of control with span of coaching: fewer managers, each holding far more people, on the theory that agents and self-directed teams absorb the coordination.

Characteristic failure

The coaching never arrives and the coordination work goes homeless. See the great flattening.

Requires, first of four

The coordination each removed layer was doing has been inventoried and reassigned before the layer goes, not after.

Intent · Composition · Motion · Proof, three cells each
LoadsStructure, Mastery
StarvesCoordination, Learning
EvidenceReportedAmazon's moves are reported; Bayer's outcomes are the company's own claims and the verdict is not in.
Shape 6

The Capability Center

Draft · sketched from research Read the full shape

Relocate ownership of an outcome to a captive unit that holds it end to end, instead of buying hours from a vendor or a shared service.

Characteristic failure

Parent retains accountability it can no longer exercise.

Requires, first of three

The parent can specify outcomes, not tasks, and has done so in writing for the work being moved.

Described in the literature Shared services organization model

Intent · Composition · Motion · Proof, three cells each
LoadsResources, Composition, Purpose
StarvesAuthority
EvidenceAdvisoryNASSCOM and equity-analyst figures; no author who has run one has written the entry yet.
Shape 7

The Levered Solo

Draft · sketched from research Read the full shape

Run a company with six to eight people and extreme revenue per head by summoning everything else: compute, distribution, support, and code.

Characteristic failure

Churn nobody puts in the deck, and no bench when one person leaves.

Requires, first of three

Product variety is small enough for eight people to hold in their heads, and the roadmap keeps it that way.

Intent · Composition · Motion · Proof, three cells each
LoadsResources, Flow
StarvesMastery, Evidence, Coordination
EvidenceSoft · company-reportedAlmost none of the revenue figures come from audited filings. The weakest evidence in the library.
Roadmap

Eleven candidates, none yet written to template

Each of these is a recognizable shape with named cases and a characteristic failure, sketched from research and not yet written up by someone who ran it. They are listed so you can argue with the sketch before the template is filled. The cells each one loads and starves are a first reading, not a finding. The sources are cited here in full because these have no library entry yet.

The Supervised Agent

Agents execute by default; humans govern by exception through review queues, approval gates, and escalation. Propose, then commit: irreversible actions wait for a recorded approval.

LoadsAuthority, Evidence, FlowStarvesMastery, Learning

Where it is seen A randomized field experiment on Taobao customer service (arXiv 2605.14830, 2026, one co-author at Alibaba) found AI-eligible chats were under 10 percent of volume and that the value of human intervention depended on the kind of AI failure and on timing. Salesforce's Omni-Supervisor monitoring (vendor material, 2025) is the productized form.

Characteristic failure Rubber-stamping. Reviewers approve blind when the queue gets noisy, and detection is delayed because the audit is asynchronous. Klarna is the cautionary case.

The structural opposite of The Hollow Core: it loads Authority and Evidence rather than starving them. First in line to be written.

Primary source A randomized field experiment with one Alibaba co-author, flagged; the Klarna reversal is independently reported; vendor material is marked as such. Candidate, not yet written to template

The Deference Ladder

A formal hierarchy for routine operations that migrates authority to whoever has the relevant expertise during an anomaly, regardless of rank, and reclaims it afterward.

LoadsMastery, Evidence, LearningStarvesFlow, Incentives

Where it is seen Aircraft carrier flight decks, commercial nuclear power, and air traffic control, the subjects of the Berkeley high reliability organization research (La Porte, Roberts, and Rochlin, 1987 to 1988), codified in Weick and Sutcliffe, Managing the Unexpected (2001).

Characteristic failure The deference reflex fails when the culture drifts toward efficiency and normalizes weak signals. Fukushima Daiichi (2011) is the standard counterexample.

The inverse of The Front-Line Command: authority is lent to expertise conditionally, not devolved permanently.

Primary source Decades of peer-reviewed research and government safety literature. Candidate, not yet written to template

The Commitment Web

No managers. Each person writes a personal mission and negotiates peer-to-peer commitments with the colleagues up and down stream. Accountability runs through those commitments and transparent metrics.

LoadsCoordination, Structure, PurposeStarvesAuthority, Mastery

Where it is seen Morning Star, where every colleague writes a Colleague Letter of Understanding, roughly 3,000 formal relationships by Gary Hamel's count (Harvard Business Review, December 2011). Buurtzorg's self-managing nurse teams with a thin coaching layer.

Characteristic failure Onboarding attrition. A Morning Star co-founder estimated about half of experienced hires leave within two years, and coordination cost reappears as negotiation overhead.

Not The Microenterprise Mesh: no internal pricing and no profit-and-loss units, only individuals bound by letters.

Primary source Independent case studies in HBR, the Commonwealth Fund, and peer-reviewed nursing literature; Corporate Rebels material is advocacy-leaning. Candidate, not yet written to template

The Assembled Crew

A whole temporary organization is cast per project from a freelance labor market, runs on well-understood trade roles for the duration, then dissolves.

LoadsComposition, Flow, ResourcesStarvesLearning, Purpose

Where it is seen Hollywood film crews, the model Adam Davidson named in The New York Times Magazine (2015). Reported spreading to cosmetics product teams and vaccine development.

Characteristic failure Failure teaches nothing. Rajaram and Muthulingam (Management Science, 2026) found that in fluid film teams failure compounds rather than improves outcomes, because the crew disbands before the verdict arrives.

The opposite of The Levered Solo in scale and permanence: large, labor-heavy, and gone at the end of each cycle.

Primary source Peer-reviewed Management Science research plus independent reporting. Candidate, not yet written to template

The Maintainer Trust

Contributions flow upward through earned trust. Maintainers gate what merges, a benevolent dictator resolves disputes and sets taste, and authority is reputational rather than appointed.

LoadsMastery, Evidence, PurposeStarvesResources, Incentives

Where it is seen The Linux kernel under Linus Torvalds and roughly a dozen subsystem lieutenants. Python under Guido van Rossum until his 2018 resignation as benevolent dictator for life.

Characteristic failure Maintainer burnout and bottlenecking. Finished code waits months to be integrated because the labor is volunteer and the apex cannot scale.

Not The Gated Pipeline: the gate is reputation, the labor is voluntary, and the slowness is a resource constraint rather than a compliance design.

Primary source Extensive primary documentation and academic study of open-source governance. Candidate, not yet written to template

The Algorithmic Crowd

A platform matches independent workers to tasks, and software does the managing: assignment, evaluation, discipline, and pay. Control runs through information asymmetry, nudges, and dynamic pricing.

LoadsCoordination, Flow, IncentivesStarvesMastery, Learning, Authority

Where it is seen Uber, the subject of Rosenblat and Stark's study of algorithmic management (working paper 2015, International Journal of Communication 2016). Deliveroo, documented ethnographically in London and Manchester. Amazon Flex.

Characteristic failure Contested legitimacy. No human holds authority, so discipline feels arbitrary, workers organize to game or resist the algorithm, and turnover and non-compliance rise.

Not The Hollow Core: the workforce is external and legally independent, and the product is a two-sided market.

Primary source A large peer-reviewed literature, including a 2025 systematic review in the Journal of Organizational Behavior. Candidate, not yet written to template

The Member Republic

A worker cooperative governed one member, one vote. A general assembly elects a council, capital is subordinated to labor, surpluses are shared and partly reinvested, and solidarity funds move labor across units in downturns.

LoadsPurpose, Incentives, StructureStarvesStrategy, Authority

Where it is seen Mondragon Corporation in the Basque Country, more than 260 businesses across 32 countries. Fagor's 2013 bankruptcy is the honest counterweight. Employee-ownership and ESOP structures are a financial cousin that conveys ownership without necessarily conveying a vote.

Characteristic failure Slow democratic decisions on capital and strategy, and tension between one-vote equality and the decisive executive authority an international group needs.

Coordinates through vote and solidarity funds where the Mesh coordinates through an internal market.

Primary source Peer-reviewed governance literature and Mondragon's own documentation. Candidate, not yet written to template

The Partner Equity

The senior producers own the firm and share its profits by seniority (lockstep) or by origination (eat what you kill), governed by partner vote with a weak central executive and a leverage pyramid of associates beneath.

LoadsIncentives, Mastery, PurposeStarvesStrategy, Coordination, Authority

Where it is seen Large law firms: Cravath left strict lockstep on December 6, 2021 (Reuters, December 7, 2021), leaving Wachtell Lipton and Debevoise as the last strict adherents. Big Four audit partnerships and McKinsey-style partner equity.

Characteristic failure Eat-what-you-kill breeds silos that suppress cross-selling and knowledge sharing, profit per partner shortens the horizon, and the managing partner cannot compel the owners.

Not the Member Republic above: ownership is concentrated among partners, voting is weighted, and associates are up or out.

Reported Independent legal and business reporting plus law-review analysis. Candidate, not yet written to template

The Buy and Build

A private-equity sponsor buys a platform company, bolts on many smaller firms in the same fragmented industry, centralizes back office and systems, keeps local operating leadership, and exits in three to five years on multiple arbitrage and debt.

LoadsResources, Strategy, IncentivesStarvesCoordination, Learning, Mastery

Where it is seen Insurance brokerage, dental, veterinary, and home-services roll-ups. Add-ons are reported at roughly three quarters of US buyout deal count; the figures come from advisory and broker sources.

Characteristic failure Integration debt and over-leverage. Industry sources put 30 to 40 percent of roll-ups below expectations on integration, multiple compression, or dis-synergies, and those sources are not independent.

Acquires external firms and centralizes them for financial reasons. Neither The Capability Center nor the Mesh.

Advisory Deal-count trends are widely reported; most quantitative detail is advisory-sourced. Held provisional until an independent or peer-reviewed source is secured. Provisional, advisory-sourced

The Scalable Command

A temporary command that expands and contracts with the incident. One incident commander holds authority, functional sections activate only as needed, span of control stays small, and roles are pre-trained and interchangeable.

LoadsAuthority, Coordination, StructureStarvesLearning, Incentives

Where it is seen The US Incident Command System inside FEMA's National Incident Management System (formalized 2004); wildfire response; hospital emergency-department command during COVID-19, documented in the emergency-medicine literature.

Characteristic failure Institutional memory dissipates when the structure dissolves, and standardized roles can suppress local knowledge.

Not The Front-Line Command: it pulls authority up to one commander for the duration, then disassembles.

Primary source Government primary documentation and peer-reviewed applications. Candidate, not yet written to template

The Focused Mission

One empowered program manager or founder is given a time-boxed budget and a mandate to force a specific capability into existence, funds or assembles the performers, and shuts the effort down when the goal is met or fails.

LoadsPurpose, Authority, ResourcesStarvesLearning, Mastery, Coordination

Where it is seen DARPA program managers, who rotate out after four to five years. Focused Research Organizations, proposed in a 2020 Federation of American Scientists report and incubated by Convergent Research and Speculative Technologies.

Characteristic failure No lasting institution. The manager rotates out, the organization is designed to end, and capability evaporates unless the transition is deliberately scaffolded.

Concentrates authority in one mission-holder for one bet with an end date. Not a pipeline and not a standing decision right.

Primary source Extensive study of DARPA; Focused Research Organizations are newer and reasonably documented. Candidate, not yet written to template
Submit a shape

Seven shapes, eleven candidates. Yours is probably missing.

A pattern is accepted when someone who has run it writes it to the template: intent, context, forces, structure, participants, testable preconditions, what it gives up, how it breaks, and known uses with sources. Two of the seven written shapes are still drafts. Contributors are named on the entry, permanently.

Draw a pattern What we are promising in return