200+ clusters diagnosed
Growing
Mature
Emerging
Transitioning
Unknown

221 clusters diagnosed · 18 countries · 9 structural patterns

ClusterOS identifies why clusters stall.
And provides the infrastructure to change it.

Diagnostic intelligence and coordination infrastructure for regional innovation ecosystems. We've diagnosed 221 clusters across 18 countries. The same 9 structural patterns appear everywhere — regardless of sector, geography or funding. ClusterOS names the pattern, finds the leverage, and gives your ecosystem the infrastructure to compound.

The methodology

Five stages.
Evidence, not opinion.

ClusterOS runs every ecosystem through a structured diagnostic pipeline — moving from raw external evidence to precise intervention points. The public pipeline observes from outside, without interviews or surveys. Commissioned diagnostics layer in actor questionnaires and steward working sessions for higher-confidence findings.

Swipe to explore →
Self-diagnostic

An interactive diagnostic journey. Select your stalls, see the reinforcing stacks, test leverage hypotheses, and configure a Digital OS architecture — all from self-assessment. The starting point, not the endpoint.

The full diagnostic

Observes from outside. Maps stalls the steward may not have named — or named incorrectly. The self-diagnostic and the commissioned diagnostic sometimes agree. Sometimes they don't. The difference is the point.

Live diagnosis

For ecosystems running the digital substrate, evidence is continuous. Every actor signal updates the diagnostic picture. The stall configuration is no longer a snapshot — it's a live read.

The platform

The diagnostic is the entry point.

ClusterOS is what comes after it — the coordination infrastructure that makes the findings actionable, keeps the ecosystem legible in real time, and turns complexity from something you periodically summarise into something you continuously govern. The diagnostic tells you what your ecosystem is doing. The platform changes it.

See how the platform works →
Who uses ClusterOS

The platform serves different actors differently. Find your entry point.

Swipe to explore →

Findings · 75 diagnostics · 18 countries

The same patterns.
Everywhere.

We have run the ClusterOS diagnostic across 75 clusters — cyber security in Belfast, Tel Aviv, Cheltenham, Singapore; advanced manufacturing in the Basque Country; the Cambridge tech ecosystem; Orlando's ten innovation clusters and more. The finding that surprised us most is not what each ecosystem is doing wrong. It is how consistently they do the same things.

"Narrative × Activity is the most frequently identified stack across all 75 diagnostics. It appears in mature ecosystems and emerging ones. Geography, sector, and maturity level do not predict its presence. Something else does."

Three stalls appear in almost every ecosystem we have diagnosed, regardless of country, sector, or maturity stage:

S2 Coordinating instead of deciding 97%
S7 Narrating instead of testing 91%
S8 Scaling activity instead of throughput 88%
S1 Re-proving instead of narrowing
S5 Mediating instead of coupling
S6 Stabilising around incumbents

Three patterns cut across everything — regardless of what a cluster makes, who funds it, or where it sits. Coordination is the default response to pressure, in every cluster we examined. The Narrative × Activity stack is the most resilient configuration we have found. And single-stall interventions almost always fail — because the stack compensates.

Full findings →
Based on your path

Continue reading

How does the diagnostic identify stalls? Who built this and why?

Framework · 5 stages · Evidence to leverage

The system isn't broken.
It's stabilising.

A stall is not a failure. It is a behavioural substitution — the system doing something observable instead of something harder. Every behaviour that looks like dysfunction makes perfect sense from inside the actor's constraints. Stalls are sensible responses that became defaults.

When stalls reinforce each other — when one lowers the cost of another — they form a stack. Stacks are why your cluster works hard and goes nowhere. And why single interventions never seem to stick.

"Every behaviour makes sense from the actor's perspective. Stalls are not failures — they are the system finding equilibria. Understanding them is the precondition for changing them."

The diagnostic moves through five stages:

01 · Evidence
What is actually happening
50–170 verified items from public records, actor questionnaires, and signal data. Confidence-tiered throughout.
02 · Patterns
What the evidence reveals
Behavioural patterns extracted from the evidence base. X-side (observable) and Y-side (what doesn't happen).
03 · Stalls
Nine canonical types
Each stall is a specific substitution pattern. Named, evidenced, and scored for intensity and confidence.
04 · Stacks
How stalls reinforce each other
The mutually reinforcing configuration — why fixing one stall fails because the others compensate.
05 · Leverage
Small perturbations that shift regimes
Not reform programmes. Not new strategies. Testable hypotheses — small withdrawals of protection that make it slightly harder for the system to continue not changing. With explicit confidence levels and defined success metrics.
Full diagnostic framework →
Based on your path

Continue reading

What did 75 diagnostics find? How is it built? Show me the architecture

Architecture · MCP substrate · Sovereign data

Intelligence lives in the protocol,
not the interface.

ClusterOS is the coordination layer for a regional economy. Not a platform with AI features added. Not a dashboard. Not a CRM with analytics bolted on. An infrastructure layer — where every actor's selfish action generates typed signals that make the whole system more intelligent, without any actor needing to care about the ecosystem.

The backend exposes the ecosystem database as a defined set of named MCP tools. The AI calls those tools at runtime to assemble what each actor needs at each moment. Intelligence is not pre-baked — it is reasoned fresh from live data on every call. When the underlying model improves, everything the platform produces improves automatically.

MCP Tool Surface · Intelligence Substrate
get_actor_profile() → role-aware profile assembly
detect_stalls() → live diagnostic against evidence base
get_matched_connections() → actor-to-actor signal matching
propagate_signal() → role-aware reframing across cluster
get_steward_diagnostic() → interrogation surface, not dashboard
get_leverage_hypotheses() → stack-calibrated intervention set
search_support_programmes() → live govt directory ingestion
ingest_evidence() → autonomous bot network write path

One event — an anchor posting a procurement RFI — produces four completely different intelligence surfaces depending on who receives it:

Actor What the AI surfaces
Founder "This RFI matches your declared capability profile. Application window: 14 days. Three preparation steps available in your journey." A specific, timely signal — not a general announcement they may never have seen.
Researcher "This procurement signal aligns with your published work. Two founders in your cluster are potential commercial intermediaries." Academic work connected to live commercial demand without monitoring procurement feeds.
Steward "Buy-side demand signal detected. Four founders have matching capability. Bridging score: 31/100. Possible Stall 4 (Anchor–Founder Disconnect)." A systems-level view, not an activity log.
Anchor "A founder in your sector cluster has reached validation stage. Capability profile now accessible. No premature disclosure required." Relevant capability surfaced at the right moment.
"Sovereign data by architecture, not policy. Each EDA operates a dedicated, isolated instance. Data does not leave the regional tenant. The EDA owns its collective intelligence permanently."
P.03
Sovereign data by architecture
One dedicated instance per EDA. Data does not leave the regional tenant. No cross-region training without consent.
P.04
Cold start solved
Autonomous bots ingest external data before any actor joins. The first founder sees a platform that already understands their ecosystem.
P.05
Interrogation, not dashboards
A dashboard answers questions someone anticipated. The steward can ask questions nobody designed for.
P.06
Continuous diagnostic
The 5-stage pipeline runs continuously as the bot network ingests signals. Not a periodic report — a live intelligence layer.
Full architecture →
Based on your path

Continue reading

Who built this? What's the background? Walk me through the diagnostic framework

About · Edinburgh · 1995 → present

We've built this kind of
infrastructure before.

In a different domain. For a different kind of complex system. The intellectual move is the same: take a system whose behaviour is hard to see, build rigorous models to make that behaviour legible, and give the institutions that govern it something they can actually act on.

"Coordination should be carried by infrastructure, not by people. When coordination is infrastructural, collaboration becomes repeatable, pathways become routable, and complexity becomes manageable rather than fragile."
1995
Barrie & Hibbert founded, Edinburgh
Andrew Barrie and John Hibbert begin as financial risk consultants. The problem: complex financial systems produce behaviours — regime switches, correlated failures, tail risks — invisible to the institutions responsible for governing them. Standard tools were inadequate. New models were needed.
1995 – 2011
Building the models
Stochastic scenario generators. Regime-switching equity models. Full-yield-curve frameworks for actuarial use. Economic scenario generators used by insurers, pension funds, and asset managers across four continents.
2011
Moody's Analytics acquires Barrie & Hibbert — $77.6M
The same intellectual move — making hidden system behaviour visible and governable — applied now to regional economies. The stewards of innovation ecosystems face the same problem financial institutions faced in 1995: consequential decisions made about systems they do not understand well enough.
2024 – present
ClusterOS
Diagnostic models to make hidden ecosystem behaviour visible and governable — for regional stewards, economic partnerships, and development agencies operating at system scale. Same intellectual move. Different complex system.
Full story →
Based on your path

Continue reading

Show me what 75 diagnostics found Request a diagnostic →

Pattern matching · Structural resemblances

Clusters with the same configuration

These clusters share your stall profile. Their diagnostic data is live — stalls, stacks, leverage hypotheses. See how the configuration plays out in different contexts.

Continue

How does the diagnostic identify these patterns? What did all 75 diagnostics find?
Your pathway
ClusterOS The Diagnostic Findings Lab Notes Get Started
Diagnose your ecosystem → See a diagnostic → Request a Diagnostic →