Fable called my idea trash. A year ago it would've called me a visionary.
I had (what I thought was) a great product idea, so I described it to Fable 5 and asked for its thoughts. It basically called the idea trash and explained why, then gave me a tweaked version that it said was better.
I thought through the tweaked version and it was actually way better than my original idea.
That personality shift matters more to me than any benchmark from the past 12 months.
A year ago, you could paste any half-baked plan into a chat window and get told it was incredible, you were brilliant, and that you should absolutely build it. Founders basically had a 24/7 yes-man for $20 a month.
Now when I bring an idea for something in @Basedash: AI data analyst the model tears it apart, explains exactly why it's weak, and then lays out what the correct version should actually look like.
Getting roasted by software stings, but as a founder, honest criticism is weirdly scarce. Your team leans agreeable, your users are polite, and investors just stop replying instead of telling you why. An AI has no social stakes, so it'll say things nobody around you will.
So yeah, the feedback feels worse now but is worth way more. The idea that got called trash came back as something genuinely better than what I walked in with (and we'll be launching it on PH soon!).
Replies
Yes, that's one of the reasons I switched from ChatGPT to Claude - it seems more honest and direct. By the way, what helps me refine the idea and make sure all nuances are thought through is "grill-with-docs" skill by Matt Pocock.
Basedash: AI data analyst
@artk nice, I'll check it out. Matt's skills have been killer.
This is such an interesting shift. I'd much rather have an AI challenge my assumptions than cheer for every idea I throw at it.
Basedash: AI data analyst
@iris_carr 100% agree
I still think founders need to sanity-check AI feedback, but having something that pushes back instead of agreeing with everything definitely makes brainstorming more useful.
The best part isn't that it said the idea was bad it's that it suggested a stronger direction. That's the kind of feedback that actually helps you build better products.
honest pushback is underrated. AI being agreeable was useful for momentum, but dangerous for judgment. The real value starts when it can find the weak assumption before you spend weeks building around it.
Basedash: AI data analyst
@nipuntaneja agreed. It gets you to slow down and ship things that actually matter, rather than every idea you have.
I agree, honest pushback is great, and actually one of my defaults in all projects in Claude is to make sure that it doesn't just agree and that responses are challenging. I also like to observe outputs from Claude with other models, for example OpenAI models, so there's also feedback from different models into the core model that I'm using as well to give a more holistic view.