I removed all AI features from my SaaS 24 hours after launch. Here is why.
I launched my product, CastIChing, a few days ago. It is a digital version of the I Ching (an ancient Chinese book of wisdom).
Like everyone else building in 2025, I thought: "I must use AI."
So, I built it. I added an AI chatbot to interpret the results. It looked cool. It felt modern.
But 24 hours after launch, I realized a big problem: It felt soulless.
The AI answers were long, but "empty."
Sometimes, it made things up (hallucinations).
For a tool about truth and fate, this was unacceptable.
So, I made a crazy decision. I deleted the AI code. All of it.
Instead of a "smart" black box, I went back to pure math. I implemented an algorithm (Box-Muller transform) to strictly mimic the traditional probability of the coin toss method.
Now, the app is simple. No chat. No fake advice. Just the raw, honest result. And guess what? It feels much better.
I wrote a full story about this decision and the "anti-hype" philosophy behind it.
👉 You can read the full manifesto here: https://castiching.com/articles/why-no-ai
My question for you: Has anyone else here removed a "cool" feature because it actually hurt the product?
That’s a bold but honest move — removing features that felt right but didn’t create clarity usually teaches more than adding new ones.
At this point, I’ve seen teams get the most leverage by asking what user behavior actually improved once the AI was gone.
Curious — what’s the one signal you’re now watching to decide what deserves to be added back (if anything)?
Hi mrHarsh, thanks for the question. It really made me think.
To answer you: The most important signal I am watching now is "Engagement with the Slow Ritual."
Here is what I found from my data:
My conclusion:
Users are actually looking for that "slow" pace and the authentic charm of the I Ching. They want the ritual, not a quick AI answer.
I do have a Reddit discussion link on the site. No feedback there yet, but the user behavior (the clicks) speaks louder than words.
That’s a really strong insight — especially the fact that users complete a deliberately slow, high-effort flow.
What stands out to me is that the “Slow Ritual” isn’t just tolerated — it’s the product. The friction is doing work, not getting in the way.
One useful way to frame this going forward might be:
Which parts of the ritual users would be disappointed to lose vs which parts they’d happily skip?
That distinction usually tells you what’s sacred vs what’s just accidental friction — and protects you from re-adding “AI speed” in places where it would actually break the experience.
Exactly. To me, the 'instant result' is what users would happily skip in other apps.
Most online tools are just 'black boxes' — you click, and you get an answer. There is no trust in that.
My original goal was to restore the full, complex process. I thought only a few niche users would accept it, but I was wrong. The data shows that for my users, the process IS the product. If I make it instant, it becomes meaningless.
That makes a lot of sense.
What stands out is that once the right uncertainty is resolved (usability / clarity / ritual / milestone), everything downstream accelerates without being forced.
It’s a good reminder that speed usually comes from fewer unknowns, not more activity.