# Offload

- **Event:** [Meta Global AI Developer Hackathon](https://cerebralvalley.ai/e/meta-ai-developer-hackathon)
- **When:** Oct 26, 12:00 PM – Nov 4, 9:00 PM (EST)
- **Where:** Online
- **Team:** [Jakub Roh](https://cerebralvalley.ai/u/jrohcz)
- **GitHub:** https://github.com/jrohcz/offload-hackathon
- **Demo video:** https://offload-demo-jrohs-projects.vercel.app/
- **Gallery:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery
- **Page:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2822

Live demo: https://offload-eta.vercel.app

Offload started with a familiar frustration: an unwanted item can sit unused because selling it demands more attention than it feels worth. Repeated questions, disappearing buyers and pickup coordination can be especially costly for people with ADHD or limited executive capacity. The aim is to make second-hand exchange easier to finish, not simply faster to list.

Offload is an AI-assisted workflow for selling an unwanted item or evaluating a second-hand offer without letting repetitive messages take over. In Sell, a photo becomes a reviewable listing draft: visible condition stays visible, missing facts become questions, and the owner approves before simulated buyer conversations begin. The agent turns a qualified offer into a single Yes/No decision card; the owner, not the model, decides.

Find reverses the workflow. The owner describes what they need, then manually enters a seller's message and optional photo. Offload checks whether the offer matches the requested item and attributes, and whether price, currency, working condition and pickup details are supported. Missing or conflicting information prompts a question or rejection instead of an invented agreement.

The demo exercises both flows with live Meta API calls, a labeled synthetic sample and five simulated buyer personas. It does not search marketplaces, contact sellers, publish listings, transfer money or complete a binding trade. Offload is a prototype for reducing the back-and-forth a person must handle; real-world time savings and completed-deal outcomes have not yet been measured.

Our ambition is to bring this workflow directly into Meta's Facebook Marketplace, where the items, people and conversations already are. A photo and a few owner-defined constraints should be enough for an agent to handle repetitive exchanges and bring back meaningful decisions. Offload is a standalone prototype of that experience; native Marketplace integration is the destination, not a capability we claim is already connected.

## More from Meta Global AI Developer Hackathon

- [Acoustic 4](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2819)
- [Kiprotich Dominic](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2820)
- [Seyon](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2821)
- [success](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2823)
- [Observer AI](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2824)
- [Dragos](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2825)

---

Markdown version of https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2822. Site index for agents: https://cerebralvalley.ai/llms.txt · full text: https://cerebralvalley.ai/llms-full.txt