WELCOME TO THE TOKEN-EFFICIENCY HACKATHON!

Everyone's building AI. Almost no one's building it lean. Every request an agent makes ships mountains of context to the model — and most of it is noise. You pay for it, you wait for it, and it doesn't make your output any better.

Paritok is the open-source model layer that fixes this, routing only the context that matters to the model. In a real SWE benchmark, that's 74% fewer input tokens per request — measured, not estimated. Real requests, real context, real reduction.

This hackathon challenges you to build a great AI project and make it lean — proving you can cut cost without cutting quality.

What Makes It Special

Paritok is open-source and built for the way people actually build AI today — no rewrites required. For this hackathon, everyone builds on Paritok's hosted GPU server:

• Point your project at Paritok's hosted GPU instead of running a model on your own machine. It's faster than a laptop GPU, and it's completely free for all hackers.
• It comes with a dashboard showing exactly how many tokens and how much cost you've saved — in real time.
• Because every request runs through the hosted server, your token savings are tracked automatically, so proving your results is effortless.

Either way, you get Paritok's open-source token-optimization models, docs, and community support to build projects that don't just demo well — they ship affordably.

And you won't be building alone. Join the Paritok Discord — our team is on hand throughout the hackathon for technical support and integration help, and it's where you'll meet other builders, showcase your work, and shape what Paritok does next.

• Use of Paritok
  How meaningfully does the project use Paritok to cut token usage? Strong submissions don't just add it — they show a real, measured reduction inside a genuine workflow.
• Technical Execution
  Quality of implementation, robustness, and whether the project actually works end-to-end. Does the code do what the submission claims, and is it built well enough for someone else to run and rely on?
• Originality
  How creative and novel is the approach? Go beyond wrapping Paritok around a hello-world app — build it into a real product, not a thin demo that could be any project with Paritok bolted on.
• Real-World Usefulness
  Would a real developer, team, or company running AI at scale see clear value in this? Submissions don't need to be production-ready, but they should solve a problem practitioners actually face.
• Submission Quality
  Quality of the demo video, written description, and README. A judge should be able to understand what the project does, why it matters, and find clear setup instructions to try it themselves.
• Bonus: Meaningful Open-Source Contribution
  Thoughtful feedback and contributions to Paritok — bug reports, feature ideas, documentation, or well-researched proposals, submitted as issues — will be looked on favorably.