Build something real in 48 hours. Impact Forge is a virtual code sprint for student developers. You get a weekend, a team, free compute, and a blank repo. What you do with it is up to you.
Free to enter. Open to any student 13+, anywhere in the world. No travel, no application, no cost.
What you get
Free AI compute. Our partner Featherless.ai gives every participant 25 inference credits — 300 for the winning team — with access to 30,000+ open-source models including Kimi k3 and GLM 5.2. Most student projects get stubbed out because API bills are real. Yours won't have to be.
A team, if you want one. Teams are 1–4 people. Solo is completely fine. If you don't have a group, our Discord has a team-matching channel where you post what you want to build and find people who want to build it with you.
Something to show for it. Every submission is a public repo, a demo video, and a technical writeup. Win or lose, you finish Sunday night with a real project you can put in front of a college admissions officer or an internship recruiter.
Tracks
Pick either one. There's no advantage to one over the other.
General Innovation
Full-stack web or mobile apps, developer tooling, productivity platforms, automation workflows. Anything you'd actually want to use.
Computational Research
Software applied to scientific computing, mathematical modeling, or biological/neuroscience data analysis. Good fit if you're interested in research and want to build toward it.
Prizes
1st place — YRI Fellowship research placement
Every member of the winning team (up to 4) gets a full research placement through the YRI Fellowship, including direct mentorship from Ivy League faculty. Valued at $1,500 per person, $6,000 for a full team. Plus 300 Featherless credits.
2nd place — cash prize
3rd place — cash prize
Schedule
All times Central Standard Time.
| When | What |
| Fri Aug 14, 6:00 PM | Kickoff and prompt release |
| Sat–Sun | Build |
| Sun Aug 16, 6:00 PM | Submissions close |
| Sun Aug 16, 9:00 PM | Winners announced |
How judging works
Every submission is scored out of 60 points across six items. Each item is scored 0–10 by our judging panel.
Technical Execution & Code Architecture — 20 pts
- Code Structure & Quality (10) — clean, modular architecture, real error handling, original logic rather than templates
- API & Compute Integration (10) — functional integration of Featherless.ai or third-party APIs; advanced inference pipelines score highest
Originality & Problem Solving — 10 pts
- Innovation & Approach (10) — solves a real problem with a fresh approach or a clever combination of tools
Utility, Functionality & Real-World Impact — 10 pts
- Functional Execution (10) — a working prototype that solves the problem for clear target users
Pitch, Demo & Documentation — 20 pts
- 3-Minute Video Demo (10) — concise, shows live execution and architecture
- Documentation & Setup (10) — complete README with setup steps, architecture diagram, and citations
Note that presentation and documentation are worth 20 of 60 points — the same as your code architecture. A working project with a confusing demo loses to a simpler project that explains itself well. Budget time on Sunday for the video and README; most teams don't.
What you need to submit
By Sunday 6:00 PM CST:
- Public repo — GitHub or GitLab, with a README that explains how to run it
- Demo video — under 3 minutes, Loom/YouTube/Vimeo, showing the thing working
- Technical writeup — short. What problem, what stack, what was hard.
Rules
- All core development happens during the 48-hour window. Open-source libraries, APIs, and frameworks are fully allowed — you just can't show up with the project already built.
- Max 4 people per team. Everyone claiming a prize must be on the Devpost roster.
- Broken links, empty repos, and private videos can't be judged. Check them before you submit.
Questions?
Join the Discord: https://discord.gg/n9xUjhWhPs — it's where team matching, announcements, and organizer Q&A all happen.
Or email ayaan.rustagi2010@gmail.com
Hosted on Devpost
