TLDR;
This video is a deep dive into B-made V6, a "Breakthrough method for agile AI-driven development." The creator explores its features, setup, and usage within an existing project, comparing it to Quickflow and other AI coding tools like Claude Code and Apex. He tests the workflow, focusing on bug resolution and feature implementation, and shares his personal opinion on its complexity and efficiency.
- B-made V6 is positioned as a tool to streamline agile development with AI.
- The creator tests its Quick Flow and B-made methods, evaluating their practicality.
- The review covers installation, project initialization, bug fixing, and feature implementation.
- A comparison is made between B-made and other AI coding tools like Claude Code and Apex, highlighting differences in autonomy and workflow.
- The creator shares his personal experience and overall opinion on B-made's complexity and usefulness.
Introduction [0:00]
The video addresses the numerous requests from viewers asking for a review and analysis of B-made, following its mention by Benjamin Code. The creator plans to thoroughly examine B-made, understand its purpose, set it up, and use it within an existing project. He emphasizes the importance of testing it in a real-world scenario rather than a new project.
Présentation de B-made V6 [1:30]
B-made is described as a "Breakthrough method for agile AI-driven development," targeting developers interested in agile methodologies. BMAD stands for Build More and Architect Dreams. The platform offers 21 specialized agents across four modules and 50 workflows designed to adapt to various complexities, from bug fixes to enterprise-level platform creation. The creator notes the release of V6 and highlights that the project has been ongoing since 2025. The workflow includes stages like discovery, brainstorm, research, product brief, PRD, planning, solution architecture, validation, stories, test design, and implementation, incorporating agile principles like sprints. The initial impression is that the system appears highly complex.
Analyse de la méthode Agile [3:00]
The creator expresses his initial reaction to the complexity of B-made, finding it overwhelming. He plans to test the Quick Flow and BMAD methods, skipping the enterprise version due to time constraints. The process involves using the install command to set up the agents and Workflow unit.
Installation et structure des agents [4:45]
The installation process is interactive, allowing users to select various options. This creates a BMAD folder with configurations for agents and custom settings. The structure includes agent manifests with CSV files and BMM files containing prompts that are easily reusable. These prompts, referred to as agents, are BMM agents rather than cloud agents. The file structure includes directories for agents, data, team, research, and workflow, with the workflow directory containing complete workflows with steps for tasks like creating a product brief.
Initialisation d'un projet existant [6:45]
The creator attempts to use the documentation for guidance, but encounters numerous 404 errors, making it difficult to understand the setup process. He decides to initialize the project by providing a project name within an existing project to continue its development. The system is expected to read existing PRD and architecture files to gather information, potentially skipping the need to answer numerous questions. However, the system prompts the user to choose between modifying an existing codebase or starting a new one, which the creator finds illogical given the existing code.
Comparaison Quickflow et B-made méthode [8:15]
The creator believes that the BMAD method will be more effective than Quick Flow, which he feels mirrors his usual workflow. He notes that Quick Flow modifies small files and initiates a discovery workflow. The system continues to request validations, which the creator finds excessive for a small team. He attempts to guide the system by referencing existing files, hoping it will use the information to pre-fill required data.
Résolution de bugs et interface [10:15]
The creator aims to resolve a bug related to the infinite scroll feature and a styling issue in his application. He copies a command to a new terminal and explains the vision for the fix. Due to a technical issue, he loses the initial text input and has to recover it. The system requires the creation of a PRD and architecture, which the creator finds redundant.
Workflow de planning et épiques [12:30]
The creator finds the process slow and cumbersome, contrasting it with the speed and freedom typically associated with AI tools. He shares the bug details, and the AI agrees to resolve it following the BMAD method, which includes planning and solutioning. The system creates three stories to address the bug: Toggle UI, selected display, etc. The creator notes that the tool transforms a CLI tool into another CLI tool, requiring manual input and validation at each step.
Validation des étapes de développement [15:15]
The system presents a series of prompts requiring specific options to proceed, creating a structured planning mode. It generates Epic Stories with bug IDs and descriptions, referencing multiple files. The creator questions the utility of validating each bug fix, finding the process incredibly slow and contrary to his usual AI usage. The system outputs extensive documentation in markdown format, including planning artifacts and epics.
Comparaison avec Claude Code et Apex [18:30]
After 25 minutes of planning, the creator is ready for implementation. He compares B-made to his usual workflow using Apex commands, which offer a high degree of autonomy. He notes that B-made's workflow reduces autonomy by requiring human validation, which is more suitable for less capable AI models. He contrasts this with Opus 4.5, which gained popularity for its independence and intelligence. The creator uses Apex commands with varying degrees of autonomy, including one-shot commands for quick feature completion and debug commands for bug resolution.
Formation et outils pour développeurs [21:15]
The creator promotes a masterclass and CLI tool that includes agents, commands, hooks, and configurations for using Cloud Code effectively. This training is designed for developers aiming to code more efficiently with AI.
Test du workflow Quickdev [24:15]
The creator tests the Quick Dev workflow to implement a feature that displays multiple thumbnails in the generation menu. He finds this workflow more useful and interesting. The system detects the context and proposes direct execution, which the creator approves. The system gathers context, identifies relevant files, creates mental notes, and presents a plan with a template. The Quick Dev workflow includes execution, self-check, adversarial review, and resolve findings.
Review de code et bilan final [28:15]
The system autonomously performs a self-check and initiates an adversarial code review, analyzing the code differences. The creator appreciates the concept of reviewing code changes and finds the system's ability to identify and correct a poorly named variable valuable. He concludes that B-made is too complex for most use cases but might be suitable for large enterprises. He invites viewers to share their experiences with the tool.
Conclusion et avis personnel [30:30]
The creator reflects on his experience, noting that B-made's feature splitting can add more time. He acknowledges that Opus 4.5 has improved context retrieval. He emphasizes that his opinion is based on his personal experience and invites viewers to form their own conclusions.
