Skip to main content
The Lineage 3 paradigms · 70 years

Three paradigms. Three cost-equation shifts.

Each paradigm is a deliberate response to what the previous one made expensive. DoCoDeGo is the third in that lineage.

1956 Era 1 / 3

Waterfall

Solved
Formalised investigation. Sequential phases produced predictability.
Cost it created
Investigation cost. Change became expensive to absorb.
Chaos → Predictability
2001 Era 2 / 3

Agile

Solved
Iterated. Distributed investigation across short cycles.
Cost it created
Coordination cost. Ceremony. Certification industries.
Predictive → Adaptive
2026 Era 3 / 3

DoCoDeGo D O C O D E G O

Solved
AI made investigation cheap. Intent became the highest-leverage human work.
Cost it created
We do not yet know. Honesty about that is part of the framework.
Implementation → Intent
Read the Manifesto

Read the lineage in full →  · why each shift happened, and what the third one demands of us.

The Framework

The third paradigm. Operationalised.

An open framework for engineering teams, knowledge workers, and any organisation accountable for what AI produces in their name — software, contracts, research, content. Anywhere a human commissions AI and signs the result.

DOcument
DO
The Author

Specification is the only source of truth.

If skipped

Code ships against undocumented assumptions. Weeks later, nobody knows what it was meant to do.

COmpose
CO
The Architect

Human owns the architecture; AI implements it.

If skipped

AI output ships unreviewed. Security gaps and silent scope misreadings become production surprises.

DEmonstrate
DE
The Gatekeeper

Value flows at the speed of validation, not schedules.

If skipped

Valid outputs queue for release windows. Value moves at the speed of coordination, not verification.

GOvern
GO
The Governor

Autonomy without accountability is catastrophe.

If skipped

The system diverges from approved intent one decision at a time, surfacing in incidents and audits.

Four pillars. One loop. Stand behind what was made.

Try it Live ICS preview

Paste a spec. Score it. Now.

A browser-side heuristic for the canonical ICS rubric. Conservative on purpose — a high score here means the spec at least has the visible shape of one a team can build from.

ICS score
0 / 100 red
  • Completeness0
  • Testability0
  • Unambiguity0
  • Threats0
Gaps

    Browser heuristic. The canonical rubric still requires human review. Open the full Spec Lab →

    // in the wild today we don't fake logos · this is what's actually shipping
    • docodego-starter
      reference impl · 7 apps · 7 packages
    • spec-linter
      CLI · pip · structural gate
    • ics-scorer
      CLI · pip · canonical ICS rubric
    • aar-gtr-tracker
      CLI · pip · AAR + Green-To-Red
    • this site
      docodego.com · open source · Cloudflare Pages
    • vscode-docodego
      editor extension · build from source · Marketplace soon
    • browser-extension-docodego
      GitHub/GitLab badge · build from source · Web Store soon
    • github-action-ics
      CI gate · build from source · Marketplace soon
    • mcp-docodego
      MCP server · build from source · PyPI soon

    // shipped a project with DoCoDeGo? tell us · we'll add you here · join the Discord

    The Manifesto

    Intent is the primary artefact.
    Implementation is a compiled derivative.
    Governance is earned through the loop.

    — The DoCoDeGo Axiom
    The shift

    From Agile to DoCoDeGo.

    Agile organised work around human coordination. DoCoDeGo organises work around intent clarity.

    Reason 01

    AI made investigation, specification, and refinement cheap.

    Agile's iteration discipline distributed an expensive investigation cost; that cost is now near zero.

    Reason 02

    Intent clarity is now the highest-leverage human activity.

    When AI generates implementation, the precision of the brief determines the quality of what gets built.

    The full Manifesto sets out the four values, the four pillars, and the Anti-Complexity Principle in detail. It is short. It is honest. It is the framework's primary document.

    Read the full Manifesto
    The four pillars

    A loop, not a checklist.

    Each pillar produces the input for the next. The output of the last is the input of the first.

    Deep dive →
    The engine

    IRAF

    Intent → Reasoning → Action → Feedback

    The delivery cycle model used in the DE pillar. Each cycle begins with approved intent, proceeds through agent reasoning, executes a demonstration, and returns feedback to governance and intent.

    Replaces the sprint and the release. The loop runs continuously — gates close on validation, not on calendar.

    I
    Intent

    Approved spec is the input — what the system must do, what it must not, how we know.

    R
    Reasoning

    Agent reasons over the spec. Every step is auditable; no black boxes.

    A
    Action

    Demonstration is composed and validated. Gates close on confidence, not calendar.

    F
    Feedback

    Telemetry returns to governance and to the next spec. Learning compounds; mistakes do not.

    Intent → Reasoning → Action → Feedback → back to Intent.

    The 16 statutes

    Sixteen principles. Four pillars. One discipline.

    Each principle is violable in a specific way; each violation is observable. Click any to read the principle, the detail, and the most common way teams break it.

    Read all 16 →
    Why now

    The bottleneck moved.

    Agile coordinated humans when implementation was the bottleneck. It served well — many DoCoDeGo principles inherit directly from it: small batches, fast feedback, working software, responding to change.

    But when AI handles implementation in minutes, the bottleneck shifts to validation clarity — and the ceremonies designed for one bottleneck become friction in the other.

    DoCoDeGo is not a replacement for Agile. It is the evolution Agile did not anticipate.

    Read the case for an evolution — eight specific challenges
    Try a tool

    The same spec, written two ways. See what the gauge sees.

    The ICS heuristic scores a spec across completeness, testability, unambiguity, and threat coverage. The full Spec Lab lives on the Practice page — paste your own spec; the gauge updates as you type.

    A vague spec 8 / 100
    "Add user login. Users should be able to sign in. Make it secure."
    • ▸ no acceptance criteria
    • ▸ no threat model
    • ▸ vague terms ("secure")
    • ▸ nothing measurable
    A well-formed spec 93 / 100
    INTENT — authenticate users with email + password; issue a session token.
    ACCEPTANCE — valid creds: 200 + JWT (user_id, exp); invalid: 401. Rate-limit after 5 attempts / 15 min.
    CONSTRAINTS — argon2id hashing; p95 < 200ms @ 100 RPS; no PII in logs.
    THREAT MODEL — credential stuffing; token replay; user-enumeration timing attack.
    OUT OF SCOPE — OAuth, password reset, MFA.
    • ▸ every criterion testable
    • ▸ named failure modes
    • ▸ measurable thresholds
    • ▸ scope explicit

    Composition begins when ICS reaches 60. The bad spec needs work. The good spec is ready.

    Open the Spec Lab
    Constitutional · Not a guideline

    Governance scales with autonomy.
    Ceremony does not scale with headcount.

    Any addition to the framework must not increase the net ceremony burden on a team at any maturity stage unless the increase in governance burden is proportional to demonstrated AI autonomy risk at that stage.

    — The Anti-Complexity Principle, DoCoDeGo Manifesto

    This principle exists because complexity is commercially advantageous to consultants and naturally accretes in frameworks that lack a structural constraint against it. The principle does the structural work that good intentions alone cannot do.

    The failure mode this framework replaces

    Specless generation.

    AI produces output that feels right, evaluated by subjective judgment rather than specified criteria, owned by no one accountable for the decision. It is fast — and unaccountable.

    Named sub-pattern

    Vibe coding

    Generating software through AI prompts without understanding the output. Iterating by feel.

    Named sub-pattern

    Specless analysis

    Asking AI to produce a report without defining the question, evidence standard, or non-negotiable constraints.

    Named sub-pattern

    Prompt-first composition

    Directing AI with ad-hoc prompts instead of structured specs. Output cannot be reliably regenerated or verified.

    DoCoDeGo's response

    Intent before generation. Validation before use. Governance of what was produced.

    // quiet updates

    A short note when something material ships.

    No funnel. No drip. We email you when a new statute lands, when an external case study is published, or when the framework version bumps. Average: about once a month.

    // prefer real-time? join the Discord →

    We use your email only to send updates. Unsubscribe in one click. No trackers in the emails.

    Alpha · Honest about it

    The framework is real.
    The community is forming now.

    DoCoDeGo is in Alpha. The framework is documented, the practices are battle-tested at small scale, and the next release is being shaped in public.

    If it produces anything, it should produce engineers and teams who think more clearly about what they are building and why.

    Two doors in
    Join the Discord

    Discord is where specs are debated, the framework gets sharper, and decisions land in writing. The conversation is the artefact.