It's not a UI problem, it's a positioning problem." How brutal IH feedback fixed my app.
Hey IH,
A few weeks ago, I posted about my failed Product Hunt launch (8 upvotes). I thought my problem was marketing. I thought I just launched on the wrong day.
Then, I asked this community for brutal feedback on my AI tool (Bunzee.ai). You guys didn't hold back. You absolutely roasted my UX, and it was exactly the wake-up call my "designer ego" needed.
Here are the harsh realities you pointed out, and how I spent the last few weeks frantically fixing them:
- The "Login Wall of Death" (The Trust Killer)
The Feedback: "I hit a login screen before getting any value. Trust is broken before it is built. I bounced immediately."
The Reality Check: You were right. I was gating the core value because I was desperate for emails.
The Fix (Apr 16 & Apr 23): I completely scrapped the initial login wall. Now, you can type your prompt before logging in. When you click "Start", it asks for a quick SNS login, but saves your input so the flow is uninterrupted.
The Result: Our CPA (Cost Per Acquisition) dropped by exactly 50% almost overnight.
- The "Blank Canvas" Syndrome
The Feedback: "I stared at the input box. What do I type? A market? A URL? I felt lost and almost left."
The Reality Check: I assumed founders knew exactly what to ask. I was wrong. The cognitive load was too high.
The Fix (May 6): We restructured the entry point to be explicitly clear. Instead of a blank box, users now choose their path: "I have an idea ➔ Validate it" or "I have no idea ➔ Explore opportunities". We stopped making the user guess.
- The "So What?" Output (Analysis Paralysis)
The Feedback: "The output feels like a generic AI analysis. I don't want to just read data; I want you to tell me what to do next. Should I build this or not?"
The Reality Check: We were acting like a data dashboard, but founders needed a decision engine.
The Fix (Apr 27 & May 6): We stopped just listing competitors. The AI now generates an explicit "Next Step" framework. We changed the UX so you can immediately see the actionable roadmap, not just the research.
- Speed is Trust
The Feedback: "An unresponsive UI for an already anxious user causes quiet panic."
The Fix (Apr 17): We realized a clunky app breaks the illusion of a "smart AI." We ruthlessly optimized our assets (WebP) and removed unnecessary client-side rendering. Our Lighthouse score jumped to 95+.
The Takeaway
I spent 10 years as a UI/UX designer making things look "pretty." But IH taught me that pretty doesn't matter if the user feels confused or pressured.
I’m still trying to perfect this balance. For those of you building AI tools: How do you balance giving users "enough value upfront to build trust" vs "gating the final output to capture the lead"? Where do you draw the line?
(Also, if anyone wants me to roast their onboarding flow in return for this community's help, drop your link below!
Your fixes point to a clear rule for gating:
Don’t gate before the user believes the product understands their problem.
The login wall was expensive because it asked for trust before value. The blank canvas was expensive because it made the founder do the positioning work. And the generic AI output was expensive because it gave information without a decision.
So I’d gate after the first “oh, this actually gets me” moment.
For Bunzee, that probably means letting the user enter the idea, choose the path, and see a partial but specific diagnosis before login. Not the full roadmap, but enough to prove the output is not generic.
Example:
Show:
problem clarity
target user
main risk
one recommended next step
Gate:
full roadmap
competitor breakdown
saved report
export/share
deeper validation plan
That way the user pays with login after trust is created, not before. Your 50% CPA drop from removing the first wall already suggests this is the right direction.
Gating after the "aha" moment turns a cold roadblock into a friendly handshake.
A partial diagnosis is exactly the proof of life a skeptical founder needs before committing. We updated Bunzee.ai with this logic, so give it a spin and see if the flow feels right now. I would love to hear if the first insight makes the login feel like a fair trade.
I know a couple of founders building AI tools who are constantly thinking about that exact balance. Happy to pass along some questions if you'd like to get their perspective for free.
I appreciate your thoughtful consideration. What do you think of Bunzee from your perspective? Please share your experiences.
Feels like a lot of AI onboarding problems are really anxiety-management problems.
The user isn’t just evaluating the product.
They’re evaluating whether they feel lost, stupid, or trapped while using it.
You nailed it because that empty box was basically a surprise exam for a tired founder. I had to learn that good design is mostly just emotional support for people who are afraid of breaking something. Once the fear goes away, the product finally gets a chance to shine.