Stop Asking Which Agentic Coding Methodology to Use

The debate over which AI-assisted development methodology to employ is misguided, according to the article. Multiple systems, including Superpowers, BMAD, Compound Engineering, OpenSpec, SpecKit, GSD, and Ralph, claim to streamline the process. However, there is no single solution that fits all situations. Each framework offers a unique layer of agentic development.

The central comparison lies between execution and intent. Superpowers, which operates like a demanding senior engineer, focuses on design, planning, strict test-driven development, and rigorous review. This approach is ideal for daily coding, solo development, and pure implementation tasks. On the other hand, BMAD functions as a full product organization, encompassing analysts, project managers, architects, and readiness gates. It should be utilized when product intent remains unclear or stakeholder complexity is significant.

Context retention is crucial in preventing information loss. OpenSpec and SpecKit address this issue differently. OpenSpec is optimized for change and brownfield execution, providing a durable change contract for ongoing work on existing production codebases. SpecKit, meanwhile, focuses on governance, establishing a constitution, traceability, and audit trails—making it suitable for heavily regulated environments.

The autonomy layer introduces the GSD and Ralph frameworks. GSD aims to maintain structured projects across context windows by employing requirements, roadmaps, and verification. However, it is worth noting that the original repository for GSD was archived in June 2026, so it should only be used for long, unattended runs once a stable fork becomes available.

Ralph, in contrast, functions in a continuous loop, implementing a test and repeating the process. Caution is advised, as without robust testing, Ralph may iterate on flawed decisions.

Finally, the meta-layer is represented by Compound Engineering. This approach shifts the focus from "How do we build this correctly?" to "How does this work make the next task frictionless?" Every feature delivery concludes by recording lessons for agents to reuse. This methodology is particularly beneficial for individuals who take on a more founder-like role rather than a typical individual contributor.

When deciding on the best approach, it is advised not to install all available options. Instead, orchestrate the appropriate sequence based on the specific situation. A recommended starting point is to begin with BMAD as a higher-level harness and combine it with OpenSpec to enhance the quality of implementation. Ultimately, the speaker suggests building a stack tailored to each unique scenario.