Method · Doc

DARE Glossary

Glossary — DARE Method

Canonical terms of the method. Alphabetical order.

Agentic Engineering

A software engineering discipline focused on working well with AI agents — including prompts, context, validation, security, and iteration. DARE is an Agentic Engineering methodology.

Architect (phase 2)

The phase in which the AI proposes architecture, stack, structure, contracts. Output: BLUEPRINT.md. Follows Design, precedes Review. Details →

Attempt

Each individual AI attempt within the Ralph Loop. Each attempt = 1 cycle of "implement → run gates → analyze result". Some implementations record these in DARE/EXECUTION/<task>/attempts/.

BLUEPRINT.md

A technical document produced by the Architect phase. It contains the architecture, chosen stack, folder structure, data models, API contracts, decisions with trade-offs, and a preliminary list of tasks. Immutable after Review without a formal cycle.

Checkpoint

An explicit point of human validation between phases. DARE has 3 mandatory checkpoints: DESIGN approval, BLUEPRINT approval, and review of each task after execution.

Context Engineering

The discipline of providing the right context, at the right time, in the right amount to AI agents. It includes prompts, examples, folder structure, and naming conventions. DARE materializes this in .cursorrules, skills, and templates.

DARE/

A folder created at the root of the target project (not the dare-method repo). It contains all the method's artifacts: DESIGN.md, BLUEPRINT.md, TASKS.md, EXECUTION/task-*.md, TELEMETRY.md. It is the "dossier" of what is being built.

Design (phase 1)

The phase in which the human defines the problem, requirements, success criteria, constraints, and non-goals. The AI assists with questions. Output: DESIGN.md. Precedes Architect. Details →

DESIGN.md

A requirements document produced by the Design phase. It focuses on what and why, never on how. It is the project's "north star" — a permanent reference throughout execution.

EXECUTION/

A subfolder of DARE/ that contains individual specs per task: task-001.md, task-002.md, etc. Each file is the input to the Execute phase for that specific task.

Execute (phase 4)

The phase in which the AI implements the tasks with the Ralph Loop active. Each task ends when all Validation Gates pass. Details →

Implementation

A variant of the method for a specific IDE/agent. Currently: implementations/cursor/ and implementations/antigravity/. Each implementation is self-contained — copy and use.

Ralph Loop

The self-correction cycle within the Execute phase. The AI implements → runs Validation Gates → if it fails, corrects → tries again → repeats until it passes (or stops at 6 attempts). Inspired by Ralph Wiggum from The Simpsons. Details →

Review (phase 3)

The phase in which the human validates the BLUEPRINT before spending tokens on execution. They either approve it, or send it back to Architect with feedback. The AI does not participate in this phase. Details →

Skill

A stack-specific rules / context file that the AI loads before generating code. E.g.: skill-laravel-api.mdc teaches Laravel conventions; skill-docker.mdc teaches multi-stage builds; skill-security.mdc teaches the OWASP Top 10. Skills are per-implementation:

Task

An atomic unit of execution. Each task has: context, objective, affected files, spec, Validation Gates, dependencies, and an estimate. It lives in DARE/EXECUTION/task-NNN.md.

TASKS.md

An overview of all the BLUEPRINT's tasks, with status (pending / in progress / done) and execution order based on dependencies.

TELEMETRY.md

An automatic log (generated by the AI during Execute) of tokens consumed, models used, time spent, and estimated cost. Input for /telemetry-report.

Validation Gates

Objective commands (tests, lint, type check) that must pass for a task to be considered done. Defined in the task spec. The Ralph Loop operates around them.

Vibe Coding

An informal pattern of AI development: "give me code that does X" + hope. No structure, no checkpoints. DARE exists as a structured alternative to Vibe Coding for cases where quality and auditability matter. See comparisons.

🔗 Related topics