A New Operating Model
How shared context changes the way we build product
The Starting Point
The Roman Army solved the fundamental coordination problem — how do you align thousands of people across vast distances? — by building a nested hierarchy. Eight soldiers to a tent. Ten groups to a century. Six centuries to a cohort. Ten cohorts to a legion.
That protocol has never been replaced. The Prussian General Staff. The American railroads. Taylor. The matrix org. Agile. Spotify squads. Every modern experiment has revealed the limitations of hierarchy, but none has broken the underlying constraint.
Until now.
The Problem
Product work is too serial, too lossy, and too dependent on human memory. At every handoff, context gets compressed. A rich discovery becomes a brief. A brief becomes a ticket. A ticket becomes a conversation that reconstructs what was already known.
A PM finds that customers hit with a surprise recurring charge within 60 days are 3x more likely to disengage. Sharp insight. Specific window. By the time it ships six weeks later, it's a generic notification that doesn't move the metric.
Nobody made a mistake. The context just didn't survive the chain.
The Shift
Most companies treat AI as a productivity enhancer — make the existing structure work slightly faster. The real opportunity isn't augmenting individuals. It's replacing what the hierarchy does.
Context can now be shared, persistent, and continuously available — replacing the coordination load that hierarchy was built to carry.
What Changes
PM writes. Design reacts. Engineering reacts. Legal gates at the end. Each function reconstructs intent from compressed artifacts. Meetings fill the gaps.
Shared context enters up front. Every function reasons against the same base. A single shared plan document replaces the sequential chain. Handoffs become continuations.
New Roles
The Model
An organization where shared context is structured and retrievable, repeatable routines govern key product work, and every unit of work leaves the system smarter for the next.
The same mechanism that makes interest powerful.
Architecture
Structured, tagged corpus AI can search and surface at the moment of relevance. Decisions, insights, constraints, failure modes — not a wiki nobody reads.
Recurring product work rebuilt to pull context automatically, surface assumptions, and engage every function from the start.
Continuous pressure-testing. Flags conflicts with known constraints. Tracks assumptions against evidence. Catches stale context.
The compounding engine. What worked, what failed, what changed — committed back. Every cycle starts from a higher base.
Implications
A significant portion of what managers currently do is route information. Status meetings. Alignment sessions. Priority negotiations. Context-reconstruction exercises.
If the context layer works, that energy redirects toward what actually requires human judgment: coaching, taste, ethical calls, novel situations, and trust-building that no system can automate.
The manager's job doesn't disappear. It clarifies.
Flagship Routine
Five days from signal to aligned commitment.
The Question
What does your company understand that is genuinely hard to understand, and is that understanding getting deeper every day?
— Jack Dorsey
For a large bank: transaction data across millions of customers, relationship depth across products, behavioral patterns that compound over years. The compound product organization turns that understanding into advantage.
The Pilot
Proof
How often does prior context get pulled into new sprints — and does it improve the outcome?
Target compression: signal to aligned commitment
Issues caught early vs. discovered during build
Speed is the leading indicator. Reuse is the proof.
One pod. Six weeks. Prove the loop.