Making LLM Tell the Truth About You 🧠

An LLM can't read your mind, but can it build a psychological profile of you based on what you like, if given the right data?

I conducted an experiment and wrote a small app to find out.

No complex RAG systems or vector databases. 0 dependencies. Only Vanilla JS, LocalStorage, and one old but effective mathematical concept.

If you properly organize the data collection interface and convert the data into a strict weighted graph, then feed it to a neural network with the right prompt, the neural network will see internal logical conflicts in your own preferences. Or not? 🤔

We'll break down the math behind this approach, the architecture of my MVP, prompt engineering, and the result today at 18:00 Moscow time in Tochka Sborki 👈🏻

The broadcast and recording will be available to participants. Access is arranged through the bot: https://t.me/TScompiler_bot