AI SUMMIT 2026
Highlights
- Where AI breaks: the real edge of machine capability
- A full day on what AI can do today, what it still can’t, and why that gap matters for games, science and creative work
- Poznań Congress Center
- Friday, October 23
- 11:00-20:00
Main Sponsor
Partners
Value Proposition
Last year we asked what happens when AI stops assisting and starts rivaling us. This year we go to the edge: the places where models still fail, and the people who find those places for a living. A competitive programmer who beat an OpenAI agent. A researcher who studies why language models lose at games. One day, no hype, no doom, just evidence from people who build, research and break AI.
Why This Summit
- Evidence over announcements: what holds up under testing, not in demos
- Failure as signal: where models stall, and what that tells us about the next five years
- Practical takeaways for production, teams and strategy
- Voices from competitive programming, AI research, mathematics and game production
Key Speakers
“Psyho”, Ex-OpenAI, Elite Coder
One of the world’s top competitive programmers and a former OpenAI researcher, Psyho won the AtCoder World Tour Finals 2025 against an experimental OpenAI coding agent. A year later, he sits down with Wojciech Ozimek for an open, hour-long conversation: what has changed since that win, how close the agents are now, and what a human still brings to hard problems. Audience Q&A included.
Maciej Świechowski, PhD, Chief Science Officer at NewBro
Maciej has spent years researching AI and computer graphics for games. His talk takes a critical look at a surprising blind spot: models that write code and pass exams still struggle to actually play games. Why do LLM-based approaches fail at tasks a child handles easily, and what does that say about how they plan, track state and reason? A research view on the limits of today’s models, and on why the current hype around LLM agents in games deserves a closer look.
More speakers coming soon.
Program
The Edge: What AI Can’t Do Yet
- Where models fail: games, planning, long-horizon reasoning
- Human vs agent: lessons from competitive programming
In Production: What Ships Today
- Creative and automation pipelines in game production
- Agents in real workflows, beyond the marketing promise
- Talks selected from the open call
The Next 5 Years
- Compute economics vs. capability: what scales, what stalls
- Where human judgment stays essential, and why
Who Should Attend
- Game developers and designers who need credible timelines
- Tech leaders making AI adoption decisions
- Business planners on a 3 to 5 year horizon
- Creatives using or questioning AI
- AI engineers and researchers seeking cross-disciplinary context
- Anyone tired of noise and looking for hard signal
What You’ll Learn
- Where AI is truly production-ready in gamedev, and where it isn’t
- Why benchmark wins don’t always translate into real tasks
- How to spot the limits of a model before you build on it
- Where ROI is real vs. illusory
Proof Point
“2025: one human beat an OpenAI coding agent. 2026: we ask what that edge looks like a year later.”
Main Sponsor
Sponsors
Patrons
