| Hey HN, I built SayIt (www.sayit.chat) over the weekend. It helps people prepare for conversations they’ve been avoiding — asking for a raise, a breakup, setting a boundary, confronting a family member. The problem I kept noticing: most people know what they want to say but freeze when it actually happens. They blank, say it badly, or back down entirely. How it works: You describe your situation in plain text
It generates a personalised opening script based on your relationship type and tone
It predicts how the other person might respond (with different scenarios)
It gives you lines for each possible reaction
Built with Next.js, Groq API (llama-3.3-70b), and hosted on Vercel. Free to use, no account needed. The hardest part was getting the AI to write scripts that sound like a real person talking, not a corporate memo. Still iterating on that. Would love feedback from this crowd — particularly on whether the predicted responses feel realistic. |
1. You should explain how you felt when you encountered step 1, using 'I' statements to describe your experience, such as 'I found it confusing' or 'I wasn't sure what to do next.'
2. You can then suggest the change you think would improve the process, saying something like 'I think it would be really helpful if it redirected straight to step 2.'
3. You may also want to ask for their perspective, saying 'What do you think about that idea?' or 'Is there a reason it's set up this way?' to show you value their input
---
I really like the idea of the "They might say this — tap to see how to respond". I found it a bit confusing that i insert the text on the main website and then basically see the text again (to modify) instead of going straight to step 2 or even having the step 2 options also on the main website.
But i found it generally well structured and i like the calming colors too. Neat.