Links
Badges
Forums
•2d ago
Building chatform.in: forms that ask follow-up questions when answers are vague
Hey PH I'm building chatform.in, forms that run as a short conversation instead of a wall of fields.
The problem I kept hitting: lead and intake forms either get abandoned halfway, or come back with answers like "soon", "not sure" or "idk" that sales can't use.
•2mo ago
I am a product designer turned builder to create the user research tool of my dreams, wdyt?
Hi!
I am Dario - a product designer with 20+ years of experience, and I will be launching my product fiuto.ai on Sep 1.
Fiuto was born as a personal tool, it is the tool I was looking for but could not find:
Creating studies is an iterative process that requires review and refining with a team. Current tools do not naturally support iteration: users need to move between tools copying and pasting content, images, videos and finally refine their studies manually with feedback generated in figma, slack, google docs and so forth. This is a process that can take days or even weeks, which naturally puts off teams that need to deliver at speed - especially when features can be implemented in a day.
Knowledge of product, audience and UX methods is a must when creating effective studies: current tools do not help users fill that gap, and instead target an audience of experts - missing a crowd of users that would benefit from user validation but do not have the skillset or the time to fill that gap without assistance. This is especially true now that product owners, developers, solo builders and even marketers can create their own prototypes using tools like lovable, replit, v0, etc.
Current tools force their users to pay per-seat to share their work. I think users should be able to share their work freely, platforms should encourage sharing, and remove as much friction from process as possible - with limitations only regarding possible abuse of platform - and that users data is theirs to keep.
Current tools sell AI as a paid premium add-on. It is also AI that cannot assist users end to end with full context but only intervene granularly at single entry points, with poor understanding of product and study, needing lots of baby sitting from skilled professionals. Fiuto is an agent first, using a research toolset it understands perfectly well.
Current tools are only starting to explore MCP integration, and only to support their in-app experience. Fiuto aims to provide full parity between MCP and the web app, so that users can work on their research without opening the browser. All tools available to in-app agent will eventually make their way to MCP.
Your users will rate the flattering version higher. That's exactly why you can't ship it.
A Stanford study just put numbers on AI sycophancy the more your model agrees with people, the more they trust it, and the worse their decisions get. For makers, that means your happiest metric might be your most dangerous one.
There's a study from this spring that I can't stop thinking about, because it quietly indicts the way most of us measure whether our AI product is working.