Durable Sessions AI UX vs Rowan Capital Allocation
// TL;DR
These two frameworks solve entirely different problems. If you are building or fixing an AI product's real-time streaming, chat, or agent experience, use the Christensen Durable Sessions AI UX Framework. If you are making capital allocation decisions, structuring investments, or evaluating private credit opportunities, use the Rowan Apollo Capital Allocation Framework. There is zero overlap in use case — your decision depends entirely on whether your problem is software architecture or financial strategy.
// HOW DO THEY COMPARE?
| Dimension | Christensen Durable Sessions AI UX Framework | Rowan Apollo Capital Allocation Framework |
|---|---|---|
| Best for | Engineers and product teams building AI chat/agent experiences that must survive disconnections, work across devices, and support live user control | Capital allocators, investment professionals, and entrepreneurs evaluating deals, structuring financing, or building financial services firms |
| Domain | Software architecture and AI product UX | Finance, private credit, and institutional investing |
| Complexity | Moderate — requires understanding of streaming protocols (SSE, WebSockets), pub/sub, and frontend/backend architecture | High — requires knowledge of capital structures, credit underwriting, institutional allocation buckets, and liability matching |
| Time to apply | Days to weeks for an architecture audit; weeks to months for full implementation of Durable Sessions layer | Hours for an initial deal evaluation; weeks to months for full capital structure design or firm-level strategy |
| Prerequisites | An existing or planned AI product with streaming responses; familiarity with SSE, WebSockets, or similar transports | A business or investment scenario to evaluate; understanding of credit vs equity risk; ideally a defined liability base or capital source |
| Output type | Architecture diagnosis, gap map of three foundational capabilities, redesigned streaming infrastructure with Durable Sessions | Investment thesis, capital structure with parcelled risk tranches, origination strategy, and ecosystem design |
| Creator background | Mike Christensen (Ably) — real-time infrastructure and streaming delivery for AI products | Marc Rowan (Apollo Global Management) — co-founder of one of the world's largest alternative asset managers |
| Core mental model | Decouple agents from clients via a persistent session layer (pub/sub); eliminate the Single-Connection Trap | Understand the business before the finance; find excess returns at institutional bucket intersections; match liabilities to assets |
| Key failure modes addressed | Stream lost on disconnect, SSE resume-cancel ambiguity, second-device blindness, orchestrator relay bottleneck | Heart attack risk (funding mismatch), cancer risk (bad asset accumulation), equity-only financing, missing fundamental good |
| Audience technical level | Software engineers, technical product managers, DevOps/infrastructure leads | CFOs, portfolio managers, GPs, institutional allocators, founders raising capital |
What does the Christensen Durable Sessions AI UX Framework do?
The Christensen Durable Sessions AI UX Framework diagnoses and fixes the infrastructure layer that makes AI chat and agent experiences fragile in production. It starts from a specific observation: most AI products stream responses over a single HTTP connection (typically SSE), and when that connection drops — due to a network switch, a tab close, or a mobile interruption — the response is lost forever.
The framework identifies three foundational capabilities every production AI product needs: Resilient Delivery (streams survive disconnections), Continuity Across Surfaces (sessions follow users across tabs and devices), and Live Control (users can steer, interrupt, or cancel agent work mid-generation). It then prescribes a specific architectural solution — Durable Sessions — which is a persistent, shared pub/sub layer between agents and clients. Agents write events to the session; clients subscribe to it. Neither holds a direct connection to the other.
This framework is narrowly focused on software architecture for AI products. It is the right tool when your AI assistant works in a demo but breaks when users switch networks, open a second tab, or press a stop button.
What does the Rowan Apollo Capital Allocation Framework do?
The Rowan Apollo Capital Allocation Framework codifies the investment philosophy and operational strategy of Marc Rowan, co-founder of Apollo Global Management. It provides a systematic approach to evaluating businesses, structuring financing, and allocating capital across risk tranches.
The framework's core insight is that the best risk-adjusted returns exist at the intersections of institutional allocation buckets — assets that are too private for the public equity bucket but too safe for the alternatives bucket. It introduces concepts like Clean Sheet Thinking (solving from first principles rather than iterating on convention), Heart Attack vs. Cancer Risk (diagnosing whether a structure is vulnerable to sudden funding collapse or slow asset deterioration), and the Cost-of-Liabilities model (matching low-cost long-duration liabilities with safe yield assets to generate excess return per unit of risk).
This framework is deeply financial. It is the right tool when you are structuring a $2 billion industrial facility financing, evaluating whether to expand into private credit, or deciding how to parcel risk across a capital stack.
How do they compare?
These frameworks do not compete. They operate in entirely different domains with zero functional overlap.
The Christensen framework solves a software engineering problem: how to make AI product experiences resilient, multi-surface, and controllable by decoupling the agent layer from the client layer through a persistent session substrate. Its audience is engineers and product teams. Its output is an architecture redesign.
The Rowan framework solves a capital allocation problem: how to evaluate businesses, structure investments, and build financial services firms using credit mentality, institutional bucket analysis, and principal alignment. Its audience is investors, CFOs, and fund managers. Its output is an investment thesis and capital structure.
The only thematic connection is that both frameworks emphasize first-principles diagnosis before action. Christensen insists you audit your streaming architecture against the Single-Connection Trap before redesigning. Rowan insists you understand the business fundamentals before making any credit decision. But the domains, audiences, outputs, and technical prerequisites are completely different.
Which should you choose?
The answer is unambiguous:
- Choose the Christensen Durable Sessions Framework if your problem is that your AI product's streaming experience breaks when users disconnect, switch devices, or try to control a running agent. You are an engineer or product team, and you need to fix your real-time infrastructure.
- Choose the Rowan Apollo Capital Allocation Framework if your problem is how to evaluate a business, structure a financing deal, allocate capital across risk tranches, or build a financial services firm. You are an investor, CFO, or fund manager, and you need a systematic investment methodology.
If you are building an AI-powered fintech product, you might eventually need both — Rowan's framework for your business strategy and Christensen's framework for your product's real-time architecture. But you will never face a situation where these two frameworks are substitutes for each other. Pick the one that matches your problem domain.
Can these frameworks be used together?
Yes, but only in the narrow case where you are building a financial technology product that involves AI agents. For example, if you are building an AI-powered investment research assistant, you would use the Rowan framework to inform the domain logic (how to evaluate deals, structure capital, assess risk) and the Christensen framework to ensure the product's streaming architecture is resilient and controllable. The frameworks address orthogonal concerns — business logic versus delivery infrastructure — and combining them is additive, not redundant.
// FREQUENTLY ASKED QUESTIONS
Is the Durable Sessions framework or the Apollo Capital Allocation framework better for building AI products?
The Durable Sessions framework is specifically designed for AI product architecture — making streaming resilient, multi-device, and controllable. The Apollo framework is an investment and capital allocation methodology with no direct application to AI product engineering. If you are building an AI product, use Durable Sessions.
Can I use the Rowan Apollo framework for software engineering decisions?
No. The Rowan framework is designed for capital allocation, investment structuring, and financial services strategy. It addresses questions like how to parcel risk across a capital stack or match liabilities to assets. For software architecture decisions — especially around AI streaming and real-time delivery — use the Christensen Durable Sessions framework instead.
What problem does the Christensen Durable Sessions framework solve?
It solves the problem of fragile AI product experiences caused by coupling response streams to a single client connection. When that connection drops, the stream is lost. The framework introduces a persistent session layer between agents and clients so streams survive disconnections, work across devices, and support live user control like stop buttons and steering messages.
What problem does the Rowan Apollo Capital Allocation framework solve?
It provides a systematic methodology for evaluating businesses, structuring investments, and allocating capital. It helps identify where excess returns exist (at institutional bucket intersections), how to diagnose structural risks (heart attack vs. cancer), and how to match low-cost liabilities with safe yield assets to generate sustainable risk-adjusted returns.
Do these two frameworks overlap at all?
No. They operate in completely different domains — software architecture versus financial strategy. The only shared trait is that both emphasize first-principles diagnosis before taking action. You would never choose between them for the same problem. Your domain determines which one applies.
Which framework is harder to implement?
The Rowan Capital Allocation framework is more complex overall, requiring deep knowledge of credit markets, institutional allocation, liability matching, and capital structures. The Christensen framework is more technically specific — requiring streaming architecture expertise — but narrower in scope. Both demand domain expertise, but Rowan's spans a broader and deeper financial knowledge base.
Who created these frameworks?
The Durable Sessions framework comes from Mike Christensen of Ably, a real-time infrastructure company, presented at an AI engineering conference. The Capital Allocation framework is derived from Marc Rowan, co-founder and CEO of Apollo Global Management, one of the world's largest alternative asset managers with over $600 billion in assets under management.
Could a fintech startup need both frameworks?
Yes. A fintech startup building an AI-powered investment research or advisory product could use the Rowan framework to inform its financial domain logic and business strategy, while using the Christensen framework to ensure its AI agent streaming architecture is resilient, multi-device, and supports live user control. The frameworks address orthogonal concerns and combine naturally.