Hard Conversation

Inspiration

We wanted to solve a simple problem: AI usually tells you what to say, but real people push back. Hard Conversation lets users practice difficult conversations before they happen.

What it does

It simulates high-stakes conversations with realistic AI personas that resist, challenge, and adapt to the user's responses, then provides evidence-based scores and feedback.

How we built it

We built a deterministic Conversation Simulation Engine with scenario, persona, pressure, scoring, feedback, and retry engines, using an LLM for natural conversation and analysis.

Challenges we ran into

The biggest challenge was preventing the AI from becoming too helpful or predictable while keeping conversations realistic, adaptive, safe, and consistently scored.

Accomplishments that we're proud of

We created an AI simulator that can dynamically increase or decrease conversational pressure and turn failure into a structured learning loop.

What we learned

We learned that building effective AI systems is not just about better prompts. Deterministic architecture and clear boundaries around the LLM are critical for reliable behavior.

What's next for Hard Conversation?

We plan to add voice conversations, more scenarios and personas, deeper progress analytics, personalized difficulty, and eventually real-time coaching and mastery tracking.

Built With

• expo.io
• fastify
• llmapi
• node.js
• postgresql
• reactnative
• supabase
• typescript