AI-powered NPCs that remember how you treated them
An AI-driven NPC system that gives game characters dynamic, in-character conversations shaped by the player's past actions — the winning project at Hack of Tomorrow.
- Award
- Winner
- Organiser
- Hack of Tomorrow · Tech of Tomorrow
- Category
- Hackathons
- Year
- 2025
- Duration
- 24 hours
The challenge
Characters in games usually repeat the same scripted lines no matter what the player has done, which breaks immersion fast. We wanted NPCs that stay in character yet react to the player — and to build it on autonomous AI agents rather than a single hard-coded prompt.
What we built
We built a 2D role-playing game whose NPCs hold branching conversations generated by a large language model. Each NPC is backed by a Fetch.ai uAgent that knows the character's persona and the player's past actions, so the character grows warmer or colder over time — which in turn changes prices, help and dialogue options.
The results
- ↗ Winning project at Hack of Tomorrow 2025 in Poznań
- ↗ Playable game with AI-driven characters built in 24 hours
- ↗ Relationships with NPCs that change gameplay
Key features
- 01
LLM-generated branching dialogue
- 02
Personas that stay in character
- 03
Memory of the player's past actions
- 04
NPCs served by autonomous uAgents
- Python
- Pygame
- Fetch.ai uAgents
- ASI-1
- LLMs
Let's build your product together
Tell us what you're building — raw notes are fine. In a 30-minute call you'll get honest feedback, a rough timeline and a budget range. No pressure, no sales pitch.
- ✓ Reply within 1 business day
- ✓ NDA on request
- ✓ Free estimate
