How I Turned 350+ Programming Books Into 24 Learning Roadmaps With AI

A developer used the opencode AI coding agent to recursively scan a personal library of more than 350 programming books and resources, organizing them into 24 topic-based learning roadmaps with foundations, intermediate, and advanced stages. The agent also restructured the actual folders, detected four exact duplicates by comparing file contents rather than filenames, and verified the file count before and after, leaving 348 resources organized across 24 topics with nothing lost. The developer recommends adding an approval step that previews planned file changes before an AI agent modifies files.

I had more than 350 programming books and resources sitting on my computer. Some algorithm books were loose in the root folder. A JavaScript book somehow ended up inside Docker. Arduino material was mixed with SEO resources. I also had duplicate copies of books spread across different folders. It was a mess. But the bigger problem wasn't the folder structure. I knew there were plenty of good books in there. I just had no idea what I should read first. Backend? Data Engineering? Advanced Python? C? Everything was mixed together. So instead of spending an entire Saturday moving files around by hand, I decided to see how far I could get with opencode. I started with a simple request I asked it to scan the whole folder recursively, group the files by topic, and create a learning roadmap for each topic. My original prompt was basically: Review the folder recursively, organize the files by topic, and create a roadmap for each topic — C Programming, Backend Engineering, Data Engineering, Data Analytics, and so on — in a Markdown file. I expected a cleaned-up list of books. What I got was much more useful. It created 24 different learning areas, each with its own roadmap. Instead of just grouping the books into folders, it organized them into stages: Foundations → Intermediate → Advanced Each book was placed where it made the most sense in the learning path. So if I want to learn C from scratch, I can see which book to start with. If I want to focus on Data Engineering, there's a progression for that too. If I'm preparing for technical interviews, I have a separate path. That was the point where the project stopped being about file organization. It became a way to answer a much better question: What should I study next? Then I let it reorganize the actual files Once the roadmaps looked good, I asked it to make the real folder structure match them. This was the part where I wanted to be more careful. Moving hundreds of files is one thing. Deleting the wrong file is another. So I had it: create new folders where needed; move books into the right topic; find duplicate files; compare duplicates by their actual contents, not just their filenames; verify the total file count before and after; update the roadmap with the new file paths. It found four exact duplicates I didn't even know I had. After everything was reorganized, I ended up with: 348 resources organized across 24 topics. And, more importantly, nothing got lost. The prompt I'd use today The first prompt worked, but I'd make it more specific now: Scan my programming library recursively, including subfolders and PDF/EPUB files. Classify every file by technical topic, such as C Programming, Backend Engineering, Data Engineering, Data Analytics, Frontend, DevOps, AI, and others. For each topic, create a learning roadmap divided into stages: foundations, intermediate, and advanced. Place each resource in the stage where it's most useful. Save everything in a ROADMAPS BY TOPIC.md file and include the goal of each stage. At the end, tell me what percentage of the files were successfully categorized. For the actual reorganization, I'd also add: Before moving anything, show me the planned file changes in a table. Only remove exact duplicates after comparing their contents. Never assume two files are duplicates just because their names are similar. When you're done, verify that the file count only changed because exact duplicates were removed, and update all paths in the roadmap. That extra approval step is worth having whenever an AI agent is going to make real changes to your files. What I'm doing next My next step is to create a checklist for each topic with: books I've finished; books I still need to read; notes; progress through the roadmap. That would turn the whole library into something closer to a personal learning system instead of a folder full of things I might read someday. I wrote up the full process, including the original prompts and the improved versions, here: https://bodanthebackend.com/articles/en/organizar-biblioteca-programacion-ia-opencode/ https://bodanthebackend.com/articles/en/organizar-biblioteca-programacion-ia-opencode/ I'm curious if anyone else has tried using an AI agent for something similar — not just coding, but organizing the resources you use to learn.