Code is cheap.Judgment isn't.Pick the right ideas.
Five filters for choosing app ideas worth your runway - before validation or a single vibe-coded sprint.
The short version
Cursor, Lovable, and Claude Code mean your cousin can ship an app by Sunday. That does not mean every idea deserves six months of runway. Selection comes before validation. Use five filters - domain edge, pain severity, defensibility path, distribution reach, and founder fit - to kill bad bets in days, not quarters. However, the filters are the same whether you are building solo, with a small team, or scaling an early startup.
I hear the same sentence every week: “I have three app ideas - which should I build?” Five years ago that question assumed build cost was the constraint. Today the constraint is judgment. The builder class exploded - solo founders, vibe coders (people shipping fast with AI coding tools), and small teams alike. However, the number of good ideas did not.
Our 7-step app idea validation framework assumes you already picked something worth testing. This post fills the gap upstream - how to choose before you spend validation budget or vibe code the wrong product for twelve weeks.
Five Filters Before You Validate
1. Domain edge you can defend
Do you understand the problem from the inside - operator, clinician, coach, tradie, educator - not from a weekend of ChatGPT research? Domain edge is not a moat by itself, but it is a filter. Copycat ideas from people with no skin in the game rarely survive first contact with real users.
2. Pain severe enough to pay for
Vitamins vs painkillers still applies. If users would shrug and use a free alternative when your app costs $9.99/month, the idea fails selection before validation even starts. Look for workflows where mistakes cost money, time, reputation, or compliance exposure.
3. A path to defensibility (not just features)
Can this idea compound toward data, habits users cannot easily replace, compliance, or network effects - or is it a thin wrapper on a public model? If the entire value prop fits in a prompt, note that honestly. You might still build to learn. However, do not confuse it with creating a viable company.
Six moats stress test4. Distribution you can actually reach
Who buys this and where do they already spend attention? App Store search, AI recommendations, communities, partnerships, field sales? "Build it and they will come" died years ago. If you cannot name two realistic acquisition paths before you build, pause.
Distribution first playbook5. Founder fit for a 2-3 year grind
Apps are not 90-day sprints. Even vibe-coded MVPs need iteration, support, retention fixes, and distribution persistence. Do you want to sell to this audience daily for years? If the honest answer is no, pick a different problem - or a different role in the stack.
Worked example: three ideas, five filters
A solo founder lists: (A) AI meal planner, (B) compliance checklist app for tradies, (C) creator analytics dashboard. Quick pass:
- A - AI meal planner: fails pain severity (free alternatives everywhere) and defensibility (prompt-describable). Kill or demote unless domain edge exists (clinical nutritionist with existing audience influence).
- B - tradie compliance: passes pain (fines, lost jobs), domain edge if founder knows the trade, defensibility path via workflow + compliance moats. Strongest candidate for validation.
- C - creator analytics: fails distribution (crowded category, incumbents own relationships) unless founder has significant warm creator network. Interesting, but not a first bet.
Selection done in an afternoon. Validation budget goes to B - not three parallel vibe-coded MVPs.
“Having ten app ideas and vibe coding the one that "sounds" the most exciting to you is not a great strategy :) Avoid procrastination and instead capture real unbiased validation from as many cold audiences as possible before building.”
Zinnia O'Brien, on what founders mistake for progress in the AI builder era.
Red Flags That Fail Selection Instantly
- ChatGPT suggested the idea and you have not talked to a human in that market
- Your pitch is feature parity with an incumbent plus AI
- You chose it because the tech stack is fun to build
- Your core feature already ships as a default inside ChatGPT, Claude, or Gemini (or is likely to soon)
- You are optimising for a hackathon demo, not someone actually paying
- The plan assumes viral growth with no channel hypothesis
- The moat is "we use AI" - which is not a moat
Founder Protection in the AI Era
Read the full series:
FAQ: Choosing What Deserves Your Runway
Is idea selection different from validation?
Yes. Selection is choosing which ideas deserve investing the time into validation. Whereas, Validation is gathering evidence on the idea you picked. Our 7-step app idea validation framework covers that phase. Skipping selection means you validate three mediocre ideas sequentially instead of filtering upfront which burns weeks or months even when each validation campaign is done well. Get selection right first, then validation, then distribution tests for ideas that survive filter 4.
I have three app ideas - which should I build?
Run the five filters on each idea in an afternoon - domain edge, pain severity, defensibility, distribution, founder fit. Most founders can kill at least one idea immediately. If two survive, pick the one with the strongest distribution path and defensibility story, not the one that sounds most exciting to build. The worked example in this post (meal planner vs tradie compliance vs creator analytics) shows the logic. Then validate the survivor properly - avoid vibe coding all three in parallel.
Can ChatGPT help me choose which app idea to build?
No - not as a decision-maker. ChatGPT is useful for brainstorming, competitor lists, and structuring your thinking. It cannot tell you which idea has real pain, reachable users, or a path to moats - and it will happily validate whichever idea you are most excited about. Use it to prep, then run the five filters and talk to humans in the market. Capture unbiased signals from cold audiences before you commit validation budget or a vibe-coded sprint. Our validation framework explains why AI output is a starting point - not a GO / PIVOT / NO-GO call.
Should I build multiple ideas in parallel now that coding is cheap?
Cheap builds make parallel experiments tempting. However, running five vibe-coded MVPs without selection discipline usually multiplies noise, not learning. Better to run the five filters on 5-10 ideas in a few days (including 5-10 conversations with your target market), and test reactions from cold audiences - not just friends who will say it sounds cool. Kill most, then properly validate the one or two that clearly show the most promise. Unsure which ideas deserve that spend? .
What if my idea fails the defensibility filter but I still want to build it?
Build small, on purpose, to learn. See if you can scope a wedge that could grow a moat over time, or treat it as a portfolio project. If your core feature already ships inside ChatGPT, Claude, or Gemini - or likely will soon - read our wrapper apps post on “unhobbling” before you scale spend. Our six-moat framework helps you see what you are missing and whether you are building a company or a learning exercise.
How do the filters work for solo founders vs small teams?
Same five filters - domain edge, pain severity, defensibility path, distribution reach, founder fit. Solo founders will feel filter 5 hardest (do you want to sell to this audience for years?). Whereas small teams will add a practical check - i.e. can you all agree on one idea long enough to validate properly and see the project through, or will you parallel-build three wrappers and learn nothing?
How long should idea selection take?
For most founders, 3-7 days of focused filtering before a formal validation sprint. Long enough to kill obvious bad fits. Short enough that you do not research yourself into paralysis. Then move to structured validation campaign with real demand tests - i.e. our 7-step app idea validation framework - and distribution-first tests for survivors that pass filter 4. Want help running selection and validation together? .
I already vibe coded something. Is it too late to select?
No - but be honest about sunk cost. Run the five filters on what you built. If it fails, pivot or stop before retention and acquisition spend.
That said, many founders arrive at 44Degrees with a working build that probably should never have been built in the first place - and that is fine, and fixable, if you act before you go too far.
Filter first, then run a proper validation campaign for a clear GO, PIVOT, or NO-GO call.
Consider exploring our app idea validation service too. if you want a second opinion on what you've already shipped.
Picked an idea? Now validate it properly.
Selection filters the list. Book a free strategy call and we will help you pressure-test the survivor with real evidence before you build.