Late nights, no backlog, and a wild idea.
That’s how HeyRival came to life. A lightweight AI tool that automates competitor research and turns it into a clean, shareable report. It started as a personal itch while validating SaaS ideas. Some industry espionage drama popped up recently, and I saw tons of posts on LinkedIn and X where people were trying to figure out how to compare companies and extract insights from different data points — like trends, pricing, hiring, funding, and more. That gave me the idea to generate shareable infographics with an AI system.
So I decided to build something. Fast.
This post breaks down:
- Why I built HeyRival
- What vibe coding means (and why it worked)
- The tech stack (and how AI changes the dev flow)
- Lessons learned from Stripe, Supabase, Claude, Perplexity, and more
🚀 Why HeyRival?
I needed a tool that:
- Automates competitor research
- Looks clean enough to drop in slides or Notion docs
- Doesn't cost €500 per month like enterprise intel tools
HeyRival was my answer. You input a few competitor names, and it returns a structured infographic report. Under the hood, it uses:
- Perplexity API to run deep web searches
- Claude to organize, summarize, and structure the insights
- A custom UI to package it all nicely
🎨 What Is Vibe Coding?
"Vibe coding" means building with no roadmap, no tickets, no specs.
Just momentum, intuition, and fast iteration.
It's about:
- Feeling your way through product decisions
- Prompting AI tools to handle the boilerplate
- Staying in flow instead of overplanning
I didn’t design this in Figma. I didn’t write a spec doc.
I opened VS Code after work, wrote a few prompts, and built the thing in about 7 days.
It’s the best way to move fast in 2025:
- Prompt ChatGPT or Claude to generate the PRD
- Extract TODOs from that into a task list
- Ask for DB schema suggestions
- Build small chunks, then prompt again: “Update the docs to reflect the changes we just made”
It’s less about typing and more about orchestrating the build through high-leverage prompts.
💡 What AI Did, and What I Still Had to Touch
Let’s get real.
Prompting can get you 80% of the way. But to ship a real product, you need to know your tools.
⚡ Stripe
Stripe is powerful but opinionated. Even with AI writing the code, you still need to understand:
- How Stripe Checkout vs. Billing works
- How to set up webhooks correctly
- How to handle pricing tiers, trial periods, and plan upgrades or downgrades
- How to map your Supabase users to Stripe customers reliably
It’s not just about code generation. It’s about understanding the lifecycle and making sure the AI isn’t hallucinating key details. That means reading docs, debugging edge cases, and making micro-decisions AI can't.
🛠️ Supabase
Supabase is magical, until it isn’t.
Sometimes the AI gives you a query that should work, but it doesn't. Or it assumes a table exists when it doesn't. Or it gets the Row Level Security wrong.
Knowing how to:
- Create policies manually
- Seed the database cleanly
- Inspect logs and errors
- Manage auth across frontend and backend
...is what makes it actually work.
🌟 Claude and MCP (Future Work)
Claude was great for summarizing competitor data and turning raw results into structured chunks. I haven’t yet wired in multi-agent prompting (like using MCP) to validate or chain AI calls, but that’s on my radar to reduce hallucination and increase report reliability.
🔧 Stack Overview
- Next.js (React 18)
- ShadCN + Tailwind for UI
- Supabase for auth and data
- Stripe for payments
- Perplexity API for research
- Claude for synthesis
AI gave me the foundation, but I had to tweak auth flows, database edge cases, webhook validation, and UI polish myself.
🙌 Who HeyRival Is For
- Indie hackers validating new ideas
- PMs tracking market shifts
- Founders doing pre-launch research
- Marketers doing quick GTM analysis
You drop in some names. HeyRival does the messy digging. You get a beautiful, export-ready snapshot.
✨ Final Thoughts
We’re entering a new era of building.
If you can prompt well and you know your tools deeply, you can ship products in days that would've taken teams weeks.
Vibe coding isn’t just about speed. It’s about trust — in your intuition, in your tools, and in your ability to debug the gaps AI can't fill (yet).
Check it out, and if you’re building with vibes too, shoot me a message. Always down to jam.