.png&w=256&q=75)
🤓 Latest Submissions
amneAI — A Cure for Codebase Amnesia
I built a TOEIC English prep app with AI. It worked perfectly. I shipped it. Weeks later, a friend asked me to add quizzes. I opened my own code and froze — I had no idea where the curriculum logic lived. That's codebase amnesia. And if you've ever vibe-coded a project, you know exactly what I'm talking about. 73% of new code is AI-generated. Developers generate 3x more code than they can manually review. We're building faster than ever, but we're also forgetting faster than ever. The gap between what we build and what we understand is growing every single day. amneAI fixes this. Paste a GitHub URL. IBM Bob Shell reads your entire repository — every file, every dependency, every structural pattern — and generates a plain-English codebase guide for your future self. Not a chatbot response that disappears. A persistent, shareable dashboard with 7 sections: Summary, Architecture, Data Flow, Key Files, Entry Points, Gotchas, and Dependencies. The Entry Points section told me exactly where to add those quizzes. The Gotchas section found invisible Unicode characters in another project I built — something I shipped and never knew existed. You can't Google what you don't know exists. That's the difference. Other tools in this hackathon generate documentation or onboarding guides from repos. amneAI is different — it explains your codebase to the developer who needs it most: future you. The solo dev who vibe-coded something three months ago. The freelancer handing off a project. The new hire staring at a codebase nobody documented. Built solo in 48 hours. Written entirely in IBM Bob IDE. Powered by IBM Bob Shell's native repository context — no RAG chunking, no embedding, no retrieval loss. Just Bob reading your whole codebase like a senior dev would. amneAI — a cure for codebase amnesia.
17 May 2026
APIlot: Intelligent API Navigator
Anyone who spends their days deep in API integration knows the pain of losing countless hours simply trying to parse documentation. Figuring out which endpoints actually exist, what fields are strictly required versus optional, and what exact data types to pass is a tedious, manual bottleneck that slows down development. I built APIlot to completely eliminate this friction. APIlot is an AI-powered API navigator that transforms static, overwhelming documentation into a clean, interactive workspace. You simply drop an API documentation URL into the app. From there, APIlot intelligently reads the docs, extracts every single endpoint, and dynamically generates an interactive form showing you exactly what fields you need to fill to build your payload. No more endless scrolling or command-F searching. Built for the AMD Developer Hackathon, APIlot relies on a robust, cloud-based architecture so no local GPU is required: • Intelligent document scraping is handled via Firecrawl. • The heavy lifting of schema extraction and LLM inference runs on AMD Developer Cloud (MI300X) instances. • Model serving is managed via vLLM on AMD Developer Cloud instances. • Agentic workflow orchestration is driven by n8n. • The frontend is a snappy React + Vite interface featuring a live, syntax-highlighted payload terminal. APIlot's goal is simple: let integrators read less, so they can build more.
10 May 2026
