E'lon
🌎 Hybrid Hackathon | 💻 Online Build: 3-8 November 2026 | On-site phase: 8 November 2026 📍 If approved for the on-site experience, you'll also have the opportunity to showcase your project live. Travel and accommodation will not be covered.
Talablar va baholash
Requirements
WHAT TO BUILD
A working AI agent (or multi-step AI workflow) that:
1. automates or accelerates one clearly identified business task for early-stage founders;
2. includes at least one human checkpoint, where the user reviews, challenges or validates the agent's output before any decision or external action;
3. shows its reasoning or sources where relevant, so the user can check the output rather than trust it blindly;
4. complies with the Equal Access AI Policy below.
EQUAL ACCESS AI POLICY
- Runtime models: the agent must run exclusively on models that are free to access for any participant (open-weight models, or free tiers of commercial APIs). Paid credits, paid API plans and premium subscriptions are not allowed for the agent's runtime.
- AI-assisted building: using AI tools to help write code or design the workflow is allowed, but must be declared.
- No secrets: never publish API keys or credentials in your repository, video or workflow export.
- Verification: judges may ask any team for a live demo or additional evidence. Undeclared use of paid AI, or a false declaration, leads to disqualification.
WHAT TO SUBMIT ON DEVPOST
- A project page in English (detailed documentation may be in French, with an English summary).
- A demo video of 3 minutes maximum, uploaded to YouTube or Vimeo, showing a complete run of the agent, including the human checkpoint.
- A way for judges to test the agent at no cost: deployed link, test environment, or workflow export with setup instructions.
- A public code repository or workflow export, with installation instructions.
- An architecture diagram.
- An AI Disclosure Statement: every model, provider and tool used, its role in the agent, the plan used (free tier or open-weight), and an estimate of tokens or calls per run.
- A Time-Saved Report: the task before and after the agent, with a measured or reasonably estimated time saving.
- A Human Judgment section: what the agent does, what the human decides, and why.
- A Responsible AI section: data used, privacy, known limitations, bias and failure cases.
Judging Criteria
Business value & time saved
25% Does the agent solve a real founder problem, and is the time saving credible and measured?
Human judgment by design
20% Does the agent strengthen the user's critical thinking instead of replacing it?
Agent design & reliability
20% Is the workflow coherent, robust and reproducible? How does it handle errors?
Frugality & accessibility
15% Does it make smart use of free models and resources? Could anyone reuse it?
User experience
10% Can a founder use the agent without help from the team?
Documentation & responsible AI
10% Is the project transparent, well documented and honest about its limits?