I Tried BMAD — This Is My Feedback

[elie29]

I recently used BMAD on a real feature for about 3 weeks, working across a Java backend and Angular frontend.

I combined BMAD with Superpowers for Claude Code and wanted to see how the methodology would hold up on a real, existing project rather than a small demo.

I came away with a mixed but positive view:

• What I would definitely keep
• What I would adapt
• Where BMAD added real value
• Where the process felt too heavy
• How I handled the API contract between frontend and backend
• What happened when BMAD and Superpowers overlapped
• When I would use BMAD again — and when I wouldn't

One of my biggest takeaways was that the source of truth matters enormously. If the specification is spread across several places, the AI can end up following different interpretations.

I documented the complete experience here:

https://elie29.hashnode.dev/i-tried-bmad-on-a-real-feature

I'd really like to hear from other BMAD users:

What has your experience been on real-world or brownfield projects? What would you keep, change, or skip?

[bmadcode]

Hi @elie29 - I will review this with our team - its quite thorough so I REALLY appreciate this. Also, the timing is great, a few of the items I noticed in here about to be improved drastically with v7.

If you are in the discord (link in the readme) please reach out and happy to discuss this also - feedback like this is what keeps us improving.

Thank you.

• BMad