DevReel – A virtual gym for practical software engineering challenges
LeetCode for the messy half—refactoring, concurrency, state mutation instead of pure algorithms.

This is a focused, practical take on 'LLM literacy'—the landing page teases live, sandboxed puzzles (forensics-style prompts like a React component causing an infinite render) and a certification badge you can earn. It sells a clear promise: train logic over syntax by forcing manual fixes without Copilot. What I want to see next is a public demo, broader language/stack coverage beyond React, and evidence of how challenges scale in difficulty.
Junior to mid-level frontend developers, code reviewers, engineering hires and anyone who wants to stop over-relying on AI-generated code
LeetCode for the messy half—refactoring, concurrency, state mutation instead of pure algorithms.
Revives deprecated OpenAI gym-http-api with Docker images and built-in browser monitoring views.
Treats Markdown files as a database so fresh agents don't need to read 70k token chat logs.
This is a practical, no-nonsense play: someone trained YOLOX from scratch, released MIT-licensed weights, and packaged a path toward running it on iOS. The value is procedural — dataset curation, training recipe, and an export/convert-for-iOS pipeline — but it's not a conceptual breakthrough; I'd like to see clear mAP numbers, model size and on-device latency benchmarks before recommending it for production.
Puzzle maker mode lets you build levels, unlike static matchstick apps.
JSON-LD recovery beats generic 404s when Cursor hits stale docs.