The Manifesto

Every decision in engineering and in business is a trade-off. Speed against stability. Scope against polish. Generality against focus. Coordination against velocity. There is no such thing as a free choice. Every gain names a cost, and every organization that denies one is quietly paying the other. The only honest question is which trade you actually made, and whether anyone still remembers making it.

Most engineering organizations made the same trade without naming it. They optimized for legibility over value. Legibility to leadership. Legibility to boards. Legibility to the kind of person who needs a dashboard to know whether the company is shipping. Layers of management exist not because the work demands coordination at that density, but because someone above needs the work translated into language they can present without asking follow-up questions.

This was tolerable when building was expensive. When producing software required large teams, long cycles, and deep coordination across handoff boundaries, the overhead had a reason to exist. The tax was real, and so was the complexity it nominally managed.

AI changed the cost function. Not theoretically. Not eventually. Now. The cost of producing working software has collapsed by an order of magnitude for anyone willing to use the tools seriously. A single person with domain expertise and an AI pair can produce in a day what a cross-functional team of eight produced in a sprint.

When building gets that cheap, coordination overhead becomes the dominant cost. Not compute. Not cloud bills. Not salaries. The most expensive line item in an engineering budget is the organizational structure itself: the meetings, the alignment rituals, the translation layers, the handoffs between people who could have just built the thing. AI did not create this problem. It made it impossible to ignore.

The symptoms are familiar. Headcount as a signal of seriousness. Titles inflated to match org-chart legibility. Architecture review boards staffed by people who have not touched production code in years. Calibration meetings where careers are negotiated by people who have never read the engineer's code. Strategy off-sites that produce mission statements nobody references.

None of this is catastrophic in isolation. All of it compounds. Each layer adds latency between the person who sees the problem and the person who can fix it. Each abstraction loses fidelity. By the time a signal reaches someone with the authority to act, it has been translated so many times it no longer resembles the original.

That is the problem.

The answer is compression, not downsizing. An organization that composes and decomposes around actual problems. As elastic as the infrastructure it builds. Work made legible through systems, not through layers of human interpretation.

The atomic unit is the fire team. Two people. One Expert Scaler and one Slop Cannon. The Expert Scaler holds architectural context and the judgment to know what should not be built at all. The Slop Cannon builds. Fast, creative, unafraid to ship rough if rough is what the problem needs right now. Together they are the 10x engineer split into two complementary halves.

Fire teams operate within streams of subject-matter expertise. Payments. Compliance. Onboarding. Risk. The stream provides continuity of context. The team itself is not permanently assigned. When the highest-value work moves, the fire team moves. The stream is the river. The fire team is the boat.

Work becomes legible through systems, not through layers of interpretation. Deployment frequency. SLO adherence. Error budgets. User adoption. The work speaks for itself when you build the systems to listen. That replaces the career ladder with visible, system-measured impact.

Leadership in this model does not disappear. It transforms. The job becomes taste, mentorship, problem curation, and technology unlocking. The person who knows which hill matters this week. The person whose thirty-minute conversation unlocks a domain expert who has been filing tickets for two quarters. That is a harder job than managing a Jira board. It requires judgment that cannot be automated.

Domain experts become builders. Juniors become dangerous. The people who used to sit in status meetings now sit in front of a terminal, paired with AI, shipping real work. Not because they were forced to. Because the tools finally met them where they are.

This is not theory. It is being built right now by every startup that ships with four people what an incumbent ships with forty. By every team that discovers its best quarter was the one where two people were on leave and the remaining three stopped attending coordination meetings.

The compression is here.

The question is not whether this transition is coming. The question is whether you lead it or get reorganized by someone who does.

Or just chat with the founder of SCI.