The SAE Manifesto

The SAE Manifesto

Status: Complete · The signable declaration at the values core of the framework — clarity over effort, the spec over the code, verification over self-review, accountability over automation, honesty over hype.

Every framework that lasts has a creed underneath it — a short statement of what it values when forced to choose. The Agile Manifesto was four such choices, written when the binding constraint was change. This is ours, written for a different constraint.

When the marginal cost of generating code approaches zero, the syntax that used to dominate delivery stops being scarce — and something else becomes the bottleneck. It is not typing speed, headcount, or framework fluency. It is clarity: the ability to state precisely what should exist, why, and how you will know it is correct. Code got cheap. Clarity, verification, and accountability did not. The SAE Framework is the operating system for that world. It does not promise to make you faster. It promises to make you honest by construction — and, as a consequence, formidable.

---

What we believe — and what we therefore do

We value clarity over effort. The work is no longer producing the artifact; it is preparing the context precise enough that the right artifact is the obvious one. Therefore we treat preparation as the real labor — we measure it as Context Entropy (1–13) and refuse to start generation until the entropy is paid down.

We value the spec over the code. Code is a disposable rendering of intent; the intent is the asset. Therefore we make the spec the single source of truth, regenerate code from it, and — our delta over prior art (Spec Kit, Kiro) — we instrument the drift between spec and implementation instead of hoping they stay aligned.

We value independent verification over self-review. An author cannot see the gap between what they meant and what they wrote — that is what authorship is. Therefore the writer never judges its own work. A fresh-context Guardian whose job is to refute, not approve, performs grey-box verification before anything is trusted.

We value accountability over automation. Automation moves work; it does not move ownership. Therefore we collapse the handoff, not the headcount — coordination cost drops from O(N²) to O(1), but the cognitive cost is conserved and concentrated on one accountable human who owns every line shipped.

We value honesty over hype. A claim with no stated limit is marketing. Therefore every bold claim in this framework ships paired with a named Reality Check — the limit, the failure mode, the case where it does not apply — as a first-class feature, not a footnote.

We value discipline and measurement over speed. Speed is a side effect, not a goal; chasing it directly is how you ship confident nonsense fast. Therefore we compete on diligence you can prove — gates, evidence, and metrics — and let velocity follow.

Reality check. A manifesto is an aspiration, not a certificate. Signing it does not make a project SAE-ready, and the values above genuinely conflict in practice — "clarity over effort" can become analysis paralysis, "verification over self-review" costs real wall-clock time, and "accountability concentrated on one human" is the bus-factor problem, not a solution to it. We hold the spec as the succession plan and concede that it is an imperfect one. These are the choices we make under tension, knowing the cost. See Objections and When NOT to Use SAE.

---

The Five Non-Negotiables

The values above are how we choose. These five are where we do not choose — the creed that defines whether a workflow is SAE at all. Break one and you are doing something else.

#	Non-Negotiable	The discipline it enforces
1	The spec is the source of truth.	Code is regenerated from intent and measured against it. If the spec and the code disagree, the spec is right and the code is a bug.
2	The writer never judges its own work.	Generation and verification are kept in adversarially independent contexts. A Guardian refutes; it does not rubber-stamp.
3	Gate the irreversible.	Auth, payments, deletion, and schema drops are never auto-applied. The blast radius of an action determines the friction it earns.
4	Isolate bounded contexts.	Work happens in encapsulated vertical slices so a mistake stays local and a context stays small enough to reason about completely.
5	You own the output.	No "the AI wrote it" defense exists. The operator is accountable for every shipped line, including the ones they did not read.

These are not best practices to weigh against deadlines. They are the load-bearing walls. The first four are how the fifth becomes survivable.

---

The signature

I will pay down clarity before I generate. I will treat the spec as truth and measure its drift. I will never let the writer judge its own work. I will gate what cannot be undone, isolate what can break, and own what ships — including the limits I just named.

If that is the standard you already hold yourself to, you are not adopting the SAE Framework. You are signing something you already believe. Start here.

---

See also: First Principles · Framework at a Glance · Module 1: The Sovereign Paradigm Shifts · Module 2: Core Pillars of SAE Architecture · Module 3: The Digital Org Chart · Objections · When NOT to Use SAE · Home