AI-Native Model-Based Systems Engineering

Navigate the system,
not views.

Cairn is where AI operates through structured, reviewable changes — not chat. Every node in your system tree can be viewed through multiple lenses. Requirements, architecture, behavior, visuals, causality — all scoped to where you are.

9 Lenses

per node

3 AI Agents

per simulation

$0.04

per sim run

200ms

re-execution

System Tree

◎Autonomous Delivery RoverSYS

◇NavigationNAV

△LIDAR ArrayL

△Stereo CameraSC

△Nav ComputerNC

◇PowerPWR

△Battery PackBA

△Power DistributionPD

◇DrivetrainDRV

△Hub Motors (×4)HM

△SuspensionSU

◇CommunicationsCOM

◇Cargo ModuleCRG

◇Thermal MgmtTHM

The Problem

Powerful tools. Painful experiences.

A $2–4B market dominated by incumbents with deep capabilities but universally poor UX. Only 5% of organizations have fully transitioned. The barriers: learning curve, cost, collaboration failures.

Cameo / CATIA MagicDassault

"Gratuitously complex UI"

UX: PoorAI: None

RhapsodyIBM

"Unintuitive and dated"

UX: DatedAI: AI Hub v1.1

Enterprise ArchitectSPARX

"Frustrating"

UX: DatedAI: None

CapellaEclipse

Method-guided, but no AI

UX: GoodAI: None

CairnAI-native

Structured AI operations, not chat

UX: ModernAI: Native

The Lens Paradigm

One node, infinite perspectives.

Click any node in the system tree and switch between lenses. Requirements, architecture, behavior, verification, visuals, causality — all scoped to where you are.

Delivery Rover/NavigationSUB-NAV

Requirements

5 traced

Interfaces

12 signals

Quality

87%

1 warning

△ LIDAR Array2 reqs

△ Stereo Camera1 req

△ Nav Computer3 reqs

AI Agent Pipeline

Structured operations, not conversations.

Every AI action flows through a five-stage pipeline. Six domain specialists — each with focused context. The output is a ChangeSet you review operation-by-operation.

⌘K Command Palette

▎

1

RouterHaiku

Classify intent → choose specialist

2

Context AssemblyCode

Fetch only what's needed

3

SpecialistSonnet

Domain expert execution

4

ValidatorHybrid

Consistency & quality check

5

User ReviewUI

Accept, reject, or cherry-pick

Change ReviewArchitect

7 ops

createNodeCooling Loop Assembly

createNodeHeat Pipe Network

createNodeActive Fan Array

createInterfaceCooling Loop → Power Bus

createRequirementThermal dissipation ≥ 85W

createRequirementRange −20°C to +55°C

createRequirementCold-start below −10°C

Visuals and 3D

See your system before it exists.

Generate 2D concept art via Gemini with six style kits. Create interactive 3D meshes — sandboxed JavaScript generates geometry via MeshBuilder, rendered through Three.js. Export as glTF.

Style Kits

Photorealistic

Blueprint

Concept Art

Clay Render

Isometric

Exploded View

3D Mesh Pipeline

1Description + concept image → Claude Vision

2MeshBuilder code (~200-400 lines)

3Static analysis (security sandbox)

4In-browser execution (~50ms)

5Three.js viewer + orbit controls

6Export as glTF 2.0

Simulation Agent

AI generates the simulation. You tweak and re-run.

Three agents read your model — extract parameters, generate Python, interpret results against requirements. Edit any parameter and re-execute in 200ms. Zero AI calls on re-run.

Endurance

4.2 hr

✓ REQ-008

Avg Power

968 W

62% motors

Charge Time

115 min

⚠ REQ-009

Deliveries

8/shift

2 charges

Pipeline

1Planner — extract params, build profile

2Script Gen — Python with numpy

3Pyodide — in-browser execution

4Interpreter — compare vs requirements

Battery State of Charge — Mission Profile

Causality Lens

What must exist before this can be realized?

Based on Dr. Robert C. Harney's Pyramid of Causality. Cairn computes prerequisite layers from your model — capstone through knowledge foundation. Gaps surface as actionable warnings.

Delivery Rover

▾

Domain Technologies

Navigation

Power

Drivetrain

Comms

Cargo

Thermal ⚠

▾

Components

LIDAR

Stereo Cam

Nav Comp

Battery

PDU

Hub Motors

Susp.

▾

Parts

CPU

IMU

BMS

▾

Interfaces

CAN Bus

DC 48V

Ethernet

MIPI

CC/CV

▾

Knowledge

REQ-001

REQ-002

REQ-003

...

"If one or more levels are missing, the capstone is unlikely to be created."

— Harney, Technology Evaluation, Ch. 1

Gap Detection

⚠ Thermal Mgmt — 0 children

Click gap → ⌘K pre-filled with decomposition prompt

Node Maturity

productionprototypeconceptmature

Maps to Technology Readiness Levels. Color-codes the pyramid per-node.

Properties and Budgets

Domain-aware. Hierarchical. Actionable.

~30 property specs across 8 engineering domains, seeded by AI during decomposition. Mass, power, cost roll up through the tree. Budget bars show usage vs. allocation.

Mass42.6 / 50 kg (85%)

Power (peak)1185 / 1500 W (79%)

Cost (est.)$18.4K / $25K (74%)

Onboarding

From blank idea to instrumented model in minutes.

Start from scratch or pick a template. The AI Inception Interview asks architecture-shaping questions, refines your description, then auto-decomposes.

Templates

Delivery Rover

★ Full Example

Orbital Data Center

Advanced

Insulin Pump

Advanced

Home Energy

Starter

Ag Drone Fleet

Intermediate

Subsea ROV

Intermediate

Inception Flow

1

Describe

Natural-language system description

2

Interview

AI asks 3-5 architecture-shaping questions

3

Refine

Iterative improvement (confidence ≥ 0.8)

4

Decompose

Auto-generate subsystems + interfaces + brief

5

Explore

Workspace with full model ready

Market Position

Five gaps define the opportunity.

A $2–4B market growing 10–16% annually. Only 5% fully adopted. SysML v2 creates a switching window. No VC-backed AI-MBSE startups identified.

01AI-Assisted DecompositionUnoccupiedZero production toolsVery High

02Behavioral Model GenerationUnoccupiedNo tool generates state machines from NLVery High

03Structured AI OperationsUnexploredChangeSets for review — no precedentHigh

04Modern MBSE UXUnmetNo VS Code / Figma-like experienceHigh

05Traceability AutomationNascentReq to arch to ver remains manualMedium

$2–4B

Market size (10–16% CAGR)

Fully transitioned to MBSE

VC-backed AI-MBSE startups

AI-native systems engineering.
Structured. Reviewable. Yours.

The thinking phase before the formal model. From blank idea to instrumented, simulation-ready system architecture — in minutes, not months.

Navigate the system,not views.