Strategy Log

A walkthrough of the vision embedding and fine‑tuning pipeline used to train a Gemma-based MTG card vision model: image collection, parallel embedding builders, dataset conversion, LoRA adapter training, and deployment tips.

Trade show cart becomes a full collection tracker with condition, finish, location, and acquisition history, plus shareable public links.

Revolutionary card scanning using Ollama's multimodal API (Gemma 3 4B or LLaMA 3.2 Vision) to 'see' and understand card images, eliminating unreliable OCR for accurate MTG card identification.

How we transformed 29 scattered markdown files into SEO-friendly blog posts and established a blog-first documentation policy that drives organic traffic.

Fix camera access denied errors in development with HTTPS configuration for MTG card scanner testing.

Deep dive into fixing cart-to-card navigation with specific art variants. Learn how Next.js static rendering can silently ignore query parameters and how to fix it.

A deep dive into debugging a critical cart bug where oracle_id was being silently stripped, causing card images to link to wrong pages. Learn how explicit field whitelisting can hide bugs and how to prevent similar issues.

How we transformed a complex AI-powered deck builder into a clean, reliable tool by removing heavy dependencies and focusing on core functionality.

Revolutionary dedicated camera page with automatic card boundary detection and manual corner adjustment for professional-grade MTG card scanning.

New visual card positioning overlay helps users align MTG cards perfectly in the scanner, mimicking ID scanner interfaces for improved OCR accuracy and art matching.

Starting a keep-it-simple refactor of the app router, beginning with the busy card detail page.

Love Commander? These 5 board games offer the same strategic depth, social politics, and engine-building satisfaction as your favorite EDH pod.