Skip to content

framework-overview.md

Latest commit

 

History

History
165 lines (108 loc) · 3.55 KB

framework-overview.md

File metadata and controls

165 lines (108 loc) · 3.55 KB
Raw

Framework

The AI Native Software Engineering Lifecycle

This project defines a disciplined lifecycle for building software in an environment where intelligence is programmable.

It is not model-specific.
It is not tool-dependent.
It is methodology-first.

The lifecycle consists of five phases.

Each phase produces artifacts.
Each phase has a clear objective.
Each phase introduces structural constraints.

The phases are sequential — but the system is cyclical.

Phase 1 — Identify the Problem

Objective: Replace enthusiasm with structured clarity.

Deliverables:

  • Precise problem statement
  • Candidate solution direction
  • Defined constraints
  • Chosen platform
  • Stable conceptual summary
  • Project name

Transition condition: The problem can be articulated clearly and defended under critique.

Phase 2 — Translate to a Structured Plan

Objective: Convert clarity into a phased execution roadmap.

Deliverables:

  • A reusable plan-generation prompt
  • A sharply scoped MVP definition
  • A phased roadmap
  • Explicit validation criteria per phase

Transition condition: The plan is coherent, bounded, and aligned with the original problem.

Phase 3 — Establish the Foundation

Objective: Create a controlled execution environment.

Deliverables:

  • Project workspace
  • Coding environment initialized
  • Persisted and stress-tested docs/implementation-plan.md
  • Version control initialized

Transition condition: The plan is persisted, clarified, reviewed, and structurally stable.

Phase 4 — Controlled Implementation

Objective: Execute the plan in bounded, validated increments.

Deliverables:

  • Partially implemented MVP
  • docs/progress.md
  • Human-validated commits
  • First testable system state

Transition condition: The MVP reaches a stable, testable state.

Phase 5 — Structured Improvement

Objective: Strengthen durability without expanding scope.

Deliverables:

  • docs/project-guidelines.md
  • Optional test expansion
  • AI-assisted pre-merge review discipline
  • docs/technical-debt.md
  • Technical backlog governance

Transition condition: System stability is reinforced and technical risk is visible.

Lifecycle Loop

This framework is not waterfall.

After Phase 5, the system returns to Phase 2 for the next expansion cycle.

New feature → new plan → controlled implementation → structured improvement.

Each loop increases capability.
Each improvement cycle increases durability.

Implementation builds.
Improvement hardens.

Clarity governs both.

Cross-Cutting Discipline

Context discipline applies across all phases.

Long sessions degrade coherence. Unmanaged context introduces architectural drift.

Structural defenses are defined in:

→ context-discipline.md

Context must be curated, not accumulated.

Core Artifacts

The methodology produces durable artifacts:

  • docs/implementation-plan.md
  • docs/progress.md
  • docs/project-guidelines.md
  • docs/technical-debt.md

Conversation is transient.
Artifacts are durable.

Governing Principles

Across all phases:

  • Specification precedes implementation.
  • Constraints precede generation.
  • Verification precedes merge.
  • Clarity precedes speed.
  • Ambiguity compounds.
  • Structure scales.

The cost of code decreases.
The cost of unmanaged context increases.

Engineering remains discipline.