Show HN: Slice and Dice – analyze and explore User Prompts at scale

Hi HN, we’ve recently sold our previous product and are currently building on new ideas. This one may be the most exciting and overlooked growth opportunity for AI products we’ve found:

TLDR: We’ve built a tool for analyzing and exploring user prompts — so you can actually understand how users are interacting with your AI product, and compare behavior across different segments (languages, paid vs free, etc).

If you’re used to Mixpanel / Amplitude / PostHog to analyze user behavior, you could notice how irrelevant they become when your product is just a chat box (or voice interface). That's because in the age of AI you don’t need button events — you need to analyze a large corpus of text.

To solve this, we’ve built what we call a Mixpanel for GenAI apps — an NLP tool to analyze and explore your user chats at scale.

We can already do:

1/ Multi-layer semantic clustering (see a big picture of all the topics and drill down)

2/ Filters and groups (compare usage between languages, demography, free/paid, etc)

3/ Latent space exploration

4/ Semantic search of prompts

5/ Topics and token usage breakdown

6/ (coming) Trends and audience drift over time

So you can answer questions such as:

- What’s the main use case of my app?

- What do users who pay the most do?

- What do users who spend the most time do?

- Which quiet audiences and use cases am I missing?

- How do the user patterns differ between languages?

- What are the new audiences we can appeal to?

Please check the link for the screenshots and instructions on how to start! Any feedback is appreciated (I don't say I won't cry if it's negative)