🛑 How I Stop Myself From Building Bad Ideas

#productdevelopment #indiehacking #startup #buildinpublic

How I Stop Myself From Building Bad Ideas

I have a problem. I get excited about product ideas way too fast.

I'll be showering, and suddenly — bam — a billion-dollar idea. I write a PRD, open a code editor, and start building. Two weeks later, I realize nobody actually needs it.

I needed a way to kill bad ideas before I invested time in them. So I built a filter.

What It Does

It's a simple workflow. You write your PRD like you normally would, then run it through a set of prompts with any AI assistant. The AI analyzes your idea across 25 different angles — stuff like market demand, competition, pricing, risks, technical complexity, and even things like compliance and accessibility.

At the end, you get an HTML dashboard. It looks like a professional investment memo, but you get it in about 5 minutes.

What You'll See in the Dashboard

The dashboard has everything on one page:

Overall score — a number from 0 to 100 with a verdict label
Radar chart — see all your scores at a glance
SWOT grid — strengths, weaknesses, opportunities, threats
Risk register — what could kill your idea, with severity levels
Market sizing — TAM, SAM, SOM breakdown
Competitors — how you stack up against existing solutions
Feature fit — which features actually matter
Pricing suggestions — what people might pay
And a lot more — 25 analysis modules total

Everything is in one HTML file. Open it, scroll around.

Don't think this is it. This is just for the reference and it does much more analysis. clone it and surf yourself.

Why the Scores Feel Harsh

The AI is told to be honest. Brutally honest. Most ideas score below 70. That's not a bug — it's the point.

I wanted a tool that tells me the truth, not what I want to hear. If my idea scores 30, I want to know why so I can either fix it or move on.

The prompts specifically tell the AI to:

Never sugarcoat
Never invent traction or users
Be direct about flaws
Use real market assumptions, not wishful thinking

How I Use It

I write a PRD for any idea that excites me. Takes about an hour. Then I run the analysis. If the score is low and the feedback reveals problems I can't solve, I drop the idea and move to the next one.

It saves me weeks of building things nobody wants.

If the score is decent and the feedback points to fixable problems, I know exactly what to work on first.

The Project Is Open Source

Everything is on GitHub — the prompts, the HTML template, the workflow instructions. You can use it with any AI chatbot you like. No installation needed.

GitHub: github.com/ihssmaheel-dev/prd-intelligence-skill

I still get excited about ideas. I just don't build all of them anymore.

Top comments (2)

Harjot Singh • May 31

This is the discipline that separates people who ship a thing that matters from people with 40 abandoned repos. The hardest version of "stop building bad ideas" is emotional, not analytical - you can know an idea is weak and build it anyway because building is fun and validating is scary. The filter that's worked for me: can I find someone who'll pre-commit (pay, pre-order, or give a real "yes I need this") before I write code? If nobody will, that's the answer.

There's a counterintuitive twist now that AI made building cheap: low build-cost makes it EASIER to build bad ideas, because the "well it's only a weekend" excuse is back. So the filter matters more, not less. I actually think cheap building is good here - it lets you build the smallest validating version of an idea fast and kill it fast (which is what I optimize for with Moonshift: prompt to a shipped SaaS on your own GitHub + Vercel, ~$3 flat, first run free). Fast-to-build means fast-to-disprove. Good post - what's your single sharpest "kill it" signal? Mine is "if I have to convince myself it's a good idea, it isn't."