PatchEd

We built PatchEd to answer a simple question:

What if your old university syllabus could update itself every year — just like software does?

Most people have PDFs from school that are now ancient history. PatchEd acts like an autonomous research agent that keeps academic subjects alive. It pulls new papers, tools, and breakthroughs from trusted sources and automatically generates “patch notes” showing what’s changed since your baseline year — complete with citations and a short delta learning plan.

How We Built It

PatchEd runs as a self-improving agent loop:

Apify scrapes curated, real-time academic and industry sources.
Senso indexes them, verifies provenance, and filters out low-trust data.
OpenAI analyzes differences between the user’s old syllabus and the new knowledge graph, then writes “patch notes” (🔥 shifts, 🛠️ tools, 📚 resources, 🔮 emerging trends).
A Vercel Cron agent runs yearly — autonomously generating, citing, and emailing updates without human input.

The stack: Next.js, Supabase, Vercel, Playwright PDF renderer, and Resend for email delivery.

🚀 What Inspired Us

We were inspired by the frustration of seeing how fast fields like AI evolve — your 2018 course is basically archaeology. We wanted an agent that could continuously learn, track real-world data, and take meaningful action (generate new patch notes + deliver them).

💡 What We Learned

We learned how to build an autonomous LLM workflow that stays grounded in evidence — no hallucinations allowed. Coordinating Apify → Senso → OpenAI → Email required clear contracts, embeddings, and error recovery logic, so the system could “self-heal” and keep improving.

🧱 Challenges

Designing reliable diff reasoning between old and modern topics (e.g., connecting SIFT (2010) → ViT (2020)).
Keeping outputs trustworthy— enforcing “no citation → no claim.”
Making the entire pipeline fully autonomous, so new years patch themselves without prompting.

✨ In Short

PatchEd is a living, learning agent that keeps your education current — automatically. It’s like having your own research assistant, always online, always up to date.