SED Development Workflow

A systematic approach from specification to production

SED follows a structured workflow with distinct phases, each with clear objectives and deliverables. This ensures consistency, quality, and predictability throughout the development lifecycle.

1

Preparation Phase

Craft complete, detailed specifications before development begins.

Key Activities:

  • Write Specification Storylines: Draft an overall storyline describing the complete product flow and detailed storylines for each feature (at least 100 items each)
  • Create Detailed Specifications: Convert every storyline item into blueprint-level detail, including:
    • Database design (tables, fields, indexes, foreign keys)
    • API endpoints and routing specifications
    • Function names, parameters, and return values
    • UI/UX requirements with exact measurements and colors
    • Complete testing specifications
  • Validate Specifications: AI analyzes and scores specifications from 0-100
  • Achieve Minimum Score: Specifications must score 90 or higher before development can proceed
2

Validation Phase

Ensure specifications are complete, consistent, and ready for implementation.

Validation Criteria:

  • Database Design: All tables, relationships, and constraints are fully specified
  • Business Logic: Every function, parameter, and workflow is documented
  • UI/UX Requirements: Complete design specifications with all visual details
  • Testing Plan: Unit tests, integration tests, and E2E test scenarios are defined
  • Deployment Environment: Infrastructure, hosting, and operational requirements are specified

Spec Completeness Score Example

When specifications are insufficient, AI returns a Spec Error:

SpecError: Insufficient specification to execute.
Reason: Database schema, authentication flow, or encryption detail missing.
Required Spec Score: ≥90
Current Score: 42
3

Execution Phase

AI implements precisely what the specification dictates—nothing more, nothing less.

Implementation Rules:

  • Strict Compliance: AI executes exactly what is written in the specification
  • No Interpretation: AI does not infer intent or make assumptions
  • No Improvisation: Anything outside the specification is ignored
  • Spec Synchronization: If humans intervene, those adjustments must be captured as updated specifications
4

Verification Phase

Automated testing validates implementation against specifications.

Verification Process:

  • Test Case Derivation: Test cases and expected outcomes are derived directly from the specification
  • Automated Testing: Unit tests, integration tests, and E2E tests run automatically
  • Spec Matching: Every feature is evaluated by the question: "Does it match the specification?"
  • Error Reporting: If verification fails, AI reports discrepancies and returns to the execution phase

Coverage Requirements

Minimum 80% code coverage across statement, branch, function, and line metrics. Tests must cover success paths, failure paths, edge cases, and error scenarios.

5

Deployment Phase

Human-supervised deployment ensures safe transition to production.

Deployment Process:

  • Human Involvement Required: AI can generate deployment scripts and configs, but humans must review and execute
  • Script Generation: AI creates deployment scripts, environment configurations, and CI/CD procedures based on specifications
  • Review Process: Humans review all deployment artifacts before execution
  • Staged Rollout: Deploy to staging first, validate, then proceed to production
6

Operation Phase

Continuous monitoring, maintenance, and iterative improvements.

Operational Activities:

  • Automated Monitoring: AI monitors system health, logs, and metrics according to operational specifications
  • Error Reporting: AI reports issues and recommends automated fixes based on specifications
  • Performance Tracking: Continuous monitoring of response times, resource usage, and user experience metrics
  • Iterative Updates: When new requirements appear, follow the loop: update spec → re-evaluate score → redevelop

Continuous Improvement Loop

1 Identify new requirement or issue → 2 Update specification → 3 Re-evaluate spec score → 4 Redevelop if score ≥90 → 5 Test and deploy

Workflow Summary

The SED workflow is designed to ensure that every step of development is traceable to specifications:

  1. Preparation: Create complete specifications (score ≥90)
  2. Validation: Verify specifications meet all criteria
  3. Execution: AI implements exactly as specified
  4. Verification: Automated tests validate implementation
  5. Deployment: Human-supervised rollout to production
  6. Operation: Monitor, maintain, and iterate with spec updates
← Previous: Getting Started Next: Philosophy →