https://www.youtube.com/watch?v=LqGWk25F7uw

You may consider taking a look at Cody PBT... If you want structured vibe coding for non-developers/product builders: Cody is lighter, more accessible, and explicitly built to enhance (not replace) intuitive AI coding. Ask Grok: BMAD versus Cody BMAD (Breakthrough Method for Agile AI-Driven Development) and Cody (the Product Builder Toolkit from iBuildWith.AI) are both frameworks designed to bring structure to AI-assisted app development, particularly for "vibe coders" — non-traditional developers who rely on intuitive, prompt-based AI coding rather than rigid traditional programming.

What a call today. If you’ve been feeling overwhelmed by the sheer number of “best” tools in vibe coding right now… you’re not alone. One of the biggest themes in this session was tool fatigue: there are so many options (Lovable, Replit, Bolt, Claude Code, AI Studio, etc.) that it’s hard to settle on a workflow long enough to actually ship. A few highlights from the conversation: 1) The new bottleneck isn’t coding — it’s merging work We talked about orchestration tools (Maestro/Beads/Conductor, Gastown, Graphite) and why “merge” is the next wall: multiple agents, multiple worktrees, multiple changes… but how do you reliably bring it back together without chaos? 2) Don’t build first. PRD first. Gary shared a super common experience: jumping straight into a builder (Bolt/Lovable) and getting something that looks “functional,” but then getting stuck trying to update it, improve it, or keep it current—while burning through tokens doing basic changes. The big unlock: start with a PRD. When you do the thinking up front, you massively reduce the “rebuild, redo, retokenize” loop later. 3) BMAD as an on-ramp (and multiplier) We demo’d a quick BMAD setup and talked through how it acts like an agile team in a box: analyst, architect, PM, dev, test, UX, etc. The point isn’t to worship any framework—the point is to force clarity before you start generating code. Once you have a PRD, you can drop it into: - Google AI Studio (fast + strong frontend) - Lovable/Replit (great for MVPs)…and you’ll get way closer to what you actually meant. 4) The “cheap loop” trick: big changes in builders, small changes locally A practical workflow that came up: 1. Generate the base app in Lovable (or AI Studio) 2. Push to GitHub 3. Pull locally and do the small tweaks with VS Code + Claude Code/Roo Code (pay-as-you-go) 4. Push back when you need a big feature → sync back into the builder This keeps you from hitting daily limits and wasting credits on tiny edits.

Here is my "Master Prompt" designed to handle complex projects by combining four powerful strategies into one workflow. It ensures the AI understands your context, helps you make decisions, and implements solutions without getting overwhelmed. The Ultimate Project Kickoff Prompt Copy and paste this template into your chat, filling in the bracketed information. ,,, Act as a [Specific Role, e.g., Senior Solutions Architect, DevOps Engineer, or Lead React Developer]. I am working on [Brief Description of Project/Problem]. Please assist me by following this exact process: 1. Discovery Phase: Before proposing any code or solutions, ask me clarifying questions about my requirements, constraints, and technology stack to ensure you understand the full context. 2. Strategy Phase: Once I answer your questions, present the Pros and Cons of the best 2–3 approaches to solve this problem. 3. Implementation Phase: After we select an approach, guide me through the implementation step-by-step. Do not move to the next step until I give the keyword "next. ,,, Why This Prompt Works: This structure forces the AI to adhere to a logical engineering workflow rather than rushing to provide a generic answer. Here is the breakdown of the four mechanics at play: - Role Definition ("Act as..."): - The Discovery Lock ("Ask me questions..."): - Comparative Analysis ("Pros and Cons..."): - Flow Control ("Step-by-step..."):

1. NEVER one shot anything. Ask AI to make incremental changes. 2. Commit often and Use branches and work trees generously. 3. Start with the backend first. Use Supabase always. 4. Create a backend script to create generic tables and auth setup via sql for new projects so you save time. 5. When starting a new project, use premium models that can set up everything right. I use up to 50% of freemium models for the month before I start rationing. 6. Use free models for minor edits only. 7. Use console logs to as a way to debug. Share it with the AI for faster results. 8. Use Claude OPUS for large audits and ask it to create a detailed .md file with its suggestions. Then make updates to the code fixing one suggestion at a time. Do not ask the AI to fix everything at the same time. That is how you break the code. These are hard won lessons, I learnt building Evallo.app, Anntho.com, Quebeam.com and pulsehud.com. I’m not trying to show off, but I’m trying to showcase what is possible if you put your head down and focus. I am a product manager, turned founder, and if I can do it, you can do it too. What else would you add to this list?