Skip to main content

​
The Container for Everything

A Strategy is the top-level construct in Memex AI. It represents an objective the team is trying to achieve: a product initiative, an architectural migration, a platform capability, a compliance requirement. It’s the answer to “why are we doing any of this work?”

​
What a Strategy contains

Purpose statement

The problem or opportunity, in enough depth that anyone (human or AI) can understand the motivation without a briefing.

Architectural vision

How the solution fits into the broader system, and what principles guide it.

Decisions

The non-obvious design choices that must be resolved to move forward.

Work Items

The scoped units of implementation, linked to decisions and to each other.
Strategies are the boundary that prevents context from sprawling. When an agent picks up a work item, it loads the strategy that contains it, not the entire organisational knowledge base. A team might have five active strategies. Each is self-contained: its own decisions, its own work items, its own dependency graph.

​
The promotion path

Work sometimes outgrows its container. A work item scoped as “add a caching layer” might reveal a design problem with its own market context, architectural trade-offs, and multiple sub-work-items. When this happens, the work item is promoted to its own Strategy. The link to the parent strategy is preserved, so you can always trace how a strategy was born.

​
Cross-strategy dependencies

Strategies are self-contained but not isolated. A work item in Strategy B might depend on infrastructure delivered by Strategy A. These cross-strategy links are explicit and tracked. When Strategy A’s work item ships, Strategy B’s blocked items are automatically unblocked.

​
Why this matters

Without a strategy, decisions and work items are just a flat list. The strategy provides the why that makes every decision intelligible and every work item purposeful.

Without a Strategy

“We decided to use PostgreSQL.”Arbitrary. Untraceable. Forgotten in a month.

With a Strategy

“We decided to use PostgreSQL because our discovery matching engine needs pgvector for embedding similarity, and the strategy requires sub-100ms lookups across 3,000 occupation vectors.”Traceable to a goal. Intelligible to any agent or human.
In the old world, this was an Epic. But an Epic is a label on a group of tickets. A Strategy is a living document that holds the reasoning, the open questions, and the architectural context that every agent needs before touching the code.