JacHacks is a virtual hackathon (May 15 - May19) where participants build real agentic AI applications using Jac, an AI-native programming language developed at the University of Michigan. Backed by NVIDIA and the NSF, this is a space for builders who want to go beyond the surface and create agents that actually do things.

Whether you're new to agentic AI or already deep in it, there's a track for you.

Get started

Registration is open now! The hackathon runs virtually from May 15-19, 2026. Once you're registered, join our discord to meet your fellow hackers, get resources, and stay up to date as we get closer to kickoff.

Pick a track, form a team, and start thinking about what you want to build. We'll handle the rest.

Details

• Teams: 1–4 people
• Skill Level: All levels welcome
• Cost: Free
• Format: Virtual
• Perks: Networking, mentorship, gain valuable experience, learn new skills, and real job opportunities

• Technical Execution
  Does the project work? Is the code clean, the architecture sound, and the demo stable? We're looking at overall engineering quality, from how well the pieces fit together to how gracefully it handles edge cases.
• Use of Jac and Jaseci
  How deeply does the project leverage what makes Jac unique? Projects that take full advantage of by llm(), walkers, graph-native data modeling, and single-file full-stack development will score significantly higher than those that use Jac superficially.
• Creativity and Innovation
  Is this a fresh idea? Does it approach a problem in a way we haven't seen before? We're not looking for another todo app. Show us something that makes us rethink what's possible with agentic AI.
• Presentation and Demo
  Was the 3-minute pitch clear, engaging, and well-structured? Could someone who knows nothing about your project walk away understanding what it does and why it matters?