AI didn't break agile. It exposed that we'd already replaced agile's principles with ticket-management theater. Intent brings it back.
In 2024, the cost of implementation collapsed by an order of magnitude. Code that took a sprint now takes an afternoon. But the machinery around that code — the ceremonies, the board columns, the estimation rituals — didn't collapse with it.
The bottleneck flipped. Implementation used to be the constraint — that's why Scrum optimized for it. Now the constraint is clarity: knowing what to build, why it matters, and how to verify it's done.
Ticket systems were designed when the atomic unit of work was a unit of human labor — a card representing hours someone would spend. That assumption now actively impedes the teams using them.
Artificial time-boxes designed to batch human labor. When AI completes tasks in hours, sprints become waiting rooms.
Estimated effort is meaningless when implementation cost is near-zero. Clarity of specification is the actual cost now.
\"To Do → In Progress → Done\" describes human hand-offs. Agents don't move cards. They execute specs and emit events.
A ticket is a container for ambiguity. It mixes the what, the why, the how, and the done-when into one undifferentiated blob.
The Manifesto's first value. People (and agents) doing the right work, not following the right process.
The Manifesto's fourth value. Continuous observation and rapid response — now possible at machine speed.
The Agile Manifesto never said "use Jira." It never said "estimate in points." It never said "hold a standup every morning." Those are implementations. Agile's actual principles — tight feedback loops, working software, responding to change — are more relevant now than ever.
"The best architectures, requirements, and designs emerge from self-organizing teams."
Intent starts from the question: what would you build if you designed a team operating model from scratch today — knowing that AI agents are first-class team members, implementation is nearly free, and the bottleneck is specification clarity?
The right values. Scrum and Kanban emerged as implementations.
Tools ate the methodology. Process became ceremony. Estimation became religion.
Code is cheap. Clarity is expensive. Ceremonies can't keep up.
Continuous observation. Declarative specs. Verifiable contracts. Observable execution.
Intent replaces the ceremony stack with a continuous loop and a two-plane architecture. Work flows through a stream. Ownership persists in a topology. They intersect at the Atom — an execution envelope, not a ticket.
Intent prescribes three independent layers, bidirectionally coupled. The Knowledge Engine compiles domain understanding. The Intent loop transforms understanding into specifications. Specs become running software. Each layer can be adopted independently.
The Knowledge Engine (Layer 1) is a separate product — separable from the Intent loop. Teams can compile domain knowledge without adopting the full methodology.
Intent doesn't add a layer on top of your existing tools. It replaces the conceptual model underneath them — swapping labor-unit tracking for specification-driven execution.
In Intent, the atomic unit of work isn't a ticket — it's a Spec (a declarative description of behavior) verified by Contracts (binary pass/fail assertions). Agents execute against specs. Humans author and review them. Done means the contracts pass, not that a card moved.
What the system should do — stated clearly enough for an agent to execute against. Persists independently. Versioned. Reviewed. The lasting blueprint.
Binary pass/fail. Given/When/Then. Can be a CLI command, a test suite, or an automated check. The lasting proof that what was built is what was specified.
A single pass through the work stream. References a Spec and its Contracts. Records what happened in this cycle. Closes when done. The Spec lives on.
Every signal carries a computed trust score. Higher trust = more agent autonomy. L0 (human drives) to L4 (full auto with circuit breakers). Math, not politics.
Intent is open source and dogfooding itself. Every signal, spec, and decision behind the project is captured using the methodology it describes.