An AI outdoor-discovery journal: one small photo challenge, put the phone down, Gemma judges your discovery
Offscreen is an AI-powered outdoor discovery journal with a simple ritual: each session gives you one small real-world photo challenge, you put the phone away and go outside, return with a photograph, and Gemma (open-weight) scores whether the discovery satisfies the challenge and writes a reflection — with journal and streak tracking. No feed, no infinite scroll, by design.
Author Harsh-0986 wrote on dev.to (2026-10-08): “I built Offscreen from the initial idea through implementation and submission planning with an agentic coding workflow — twenty-two commits, phase by phase, from a written specification.” It was submitted to the Hacktoberfest '26 Open-Source AI Challenge (Week 1: Touch Grass). The backend is FastAPI + Pydantic with SQLite via SQLAlchemy, the frontend is Next.js, and workflow orchestration uses LangGraph, with Gemma served through the Google AI Studio API (all author-stated). The repo (github.com/Harsh-0986/OffScreen) was pushed on 2026-10-08 and includes a 63-second real demo video.