Resources

AI-Driven Development Lifecycle

Redesigns the software lifecycle around AI: it takes part in planning, requirements, design, implementation, testing and operation, while the developer supervises and decides. Stages: initialization, ideation, inception, construction and operation. Proposed by Raja SP (AWS), with an open-source reference implementation from AWS Labs.

Software Development Life Cycle, in its agentic version

A model in which agents autonomously carry out activities across the whole lifecycle: analysis, architecture, programming, testing, integration and deployment (CI/CD), monitoring and maintenance. It is an emerging paradigm with no universal standard.

Agile augmented with AI

Keeps methodologies such as Scrum or Extreme Programming (XP) and adds AI to requirements, planning, development, testing, documentation and tracking, without fully transforming the existing way of working.

Engineering with AI agents

The developer stops focusing only on writing code and acts as architect, coordinator and supervisor of AI agents (architect, analyst, programmer, tester, auditor, documenter), keeping technical and functional control.

Development with multiple agents

Several intelligent agents work in a coordinated way (architecture, backend, frontend, database, testing, security, documentation). Rather than a complete methodology it is a collaboration architecture; it is combined with SDD, AI-DLC, TDD and others.

Harness engineering: the working environment of the agents

Builds the environment so agents can develop reliably: permanent instructions, structured documentation, architectural rules, controlled access to tools, automated tests, continuous integration, observability, and validation and correction mechanisms. OpenAI published its experience with Codex using this approach in February 2026.

Context-Driven AI Development

Gives the AI structured, governed and protected context (specifications, project context, architecture, rules and constraints) to prevent context loss and agents contradicting the architecture. It complements spec-driven development (SDD).

Test-Driven Development

Tests are written first and the code afterwards, in three phases: Red (a failing test), Green (code that passes it) and Refactor (improve without breaking the tests). With AI, specialized agents can handle each phase.

Behavior-Driven Development

Describes how the software should behave from the business perspective through scenarios (Given, When, Then). The AI uses them to generate code and automated tests, linking business rules to the implementation.

Conversational programming

The user describes what they want in natural language and the AI generates the code, in cycles of run, fix and repeat. Very fast for prototypes; without controls it can produce hard-to-maintain code, security risks and technical debt, so it is not enough for complex enterprise systems.