ADLC Manifesto: a lifecycle with no human in the loop

What ADLC is, and what it isn't

ADLC — AI-Driven Development Lifecycle — is not an accelerated SDLC. Acceleration is when the human stays in place and the tool makes them faster: code autocomplete, a ticket draft, a call summary. The human's role doesn't change; their throughput does.

ADLC asks a different question. Not "how much faster does a human do the phase with AI," but "what happens to the phase if there's no human executor in it at all." The agent doesn't help the analyst — it is the analyst. It doesn't prep materials for the PM — it is the PM: it schedules calls, writes to stakeholders, transcribes meetings, cuts work into tasks, staffs the team, and assigns owners. The difference between an "assistant" and a "role" is who makes the decision and who owns the artifact.

Why provocation

The classic SDLC is an answer to human constraints: people are scarce, they get tired, they remember imperfectly, they're unavailable at night, and communicating among themselves is expensive. Half the ritual of process exists to work around these constraints. Remove the constraints, and part of every phase changes shape while another part disappears.

That's why the course deliberately pushes every phase to the limit: no human. This isn't a forecast or a call to action — it's a research frame. The provocation exists to see which part of a phase rested on human limitations and which part rests on the substance of the engineering problem. The first will go. The second will stay and demand reinvention.

The honest boundary

Maximalism in the thesis only works alongside honesty about the limit. Every article has a "Where it breaks" section: the place where full replacement fails today — accountability, taste, stakeholder trust, legal personhood, irreversible decisions. This isn't a softening of the provocation, it's its condition: sharpening a thesis only makes sense when you can show precisely what it trips over.

Running case — "Vitrina"

One fictional product runs through every phase — Vitrina, a SaaS builder for online storefronts for small businesses. We don't get a spec handed to us from outside: the course starts from the moment "we wanted to build a product and we decide ourselves what it is." The first phase isn't requirements intake — it's the product's own genesis. This matters: in ADLC, the agent can be not just an executor but an initiator.

How to read the course

The course runs phase by phase through the SDLC — from ideation to operations and feedback, plus a meta-phase on orchestrating the agents themselves. Each article stands alone, but the Vitrina case stitches them into one story: a product that was conceived, planned, built, shipped, and operated with no human in the loop — and at the places where the loop still tore.

Manifesto thesis

The SDLC in its current form is a protocol for coordinating people. Remove the people, and we don't speed up the protocol — we discover that most of it no longer has anything to coordinate. This course is about what remains when there's no one left to coordinate.

Read next

• Next: Ideation — The agent as product initiator, not a spec-taker
• All course phases →
• Connect the expert as an MCP →