What the junk drawer is
If you scroll to the bottom of my GitHub profile you find a class of repository that has no README worth reading, no releases, no stars, no branches, no clear point. Dump. python-code-dump. Random-python-scripts. Random-c-scripts. Dump-of-drive. Ideas. Cool-websites. linkedindump. obsidian. obsidiandata. drawing. CapCut-Drafts. Sound-Recordings. OneNote-Notebooks. gyb-backup. adventofcode from June 2026 with two days of Python in it and no attempt at day 2. fonts, which is exactly what it says: a folder of .ttf and .otf files I once put on a website.
These are not portfolio pieces. They are not meant to be read. Most of them have a single commit called Add files via upload because I dragged and dropped a folder into the GitHub web UI. Some of them are still empty six months after I made them. transfer-repo is literally empty. I made it in February 2025 to move a folder from one laptop to another and never put anything in it, because by the time I remembered it, the folder was already on the new laptop.
I refuse to delete any of them. This post is about why.
The Ideas repo is the load-bearing one
The most useful junk-drawer repo I own is called Ideas and it is a single README.md that I have edited around a dozen times over three years. Each edit adds a dated section with bullets. It reads like a diary written by someone who does not want to write a diary. Sample bullets from real dates:
- "Sublime extension to write code inside leetcode side by side, run with sample cases, where each letter you type is also typed into the website at the same time. Line by line is better."
- "Bot which identifies music in small videos like shorts, called with @bot name, use YouTube notification API to pass into Shazam alternatives to get the result."
- "Map function call and draw image of flowchart from python code."
- "Track keypress and expand from that point."
- "Halloween website where cursor is the light circle."
- "Onlyfandom onlyfriends." (I do not know what past me meant.)
- "If i face some functions which like missing library and i fix it by doing something list of steps and automatically post is as a medium article."
Some of these have strike-throughs. Those are the ones I built. Most do not, and I will never build most of them. That is fine. The repo is a low-cost place to write down a shape of an idea in ten seconds so it does not evaporate.
The load-bearing property of Ideas is that it is searchable. When I sat down to build the command palette on this site, I searched the repo for search and found a 2022 note about wanting a cross-app clipboard buffer. When I built CipherStack, I found a 2022 note about auto-rotating credentials from a dashboard. Neither of those bullets was the design. But both of them were evidence that the itch was three years old, not three days old, and that mattered when I chose what to spend a weekend on. A clean folder cannot tell me that.
Dumps of what was on my laptop
python-code-dump (September 2022) contains files with names like 1662449203437.pdf, Screenshot_2022-07-23-15-03-27-166_com.android.thememanager.jpg, from itertools import permutations.py, and Leetcode.java. It is 29 MB. It is what happens when I sit at my desk before wiping a laptop and drag everything out of a folder called code into a GitHub repo because I do not trust that I will not need one of those files someday.
Dump-of-drive (December 2022) is the same operation but for a Google Drive migration. It has a shortcut to a network drive, a copy of Sublime Text 3211, a .gitconfig, and homework PDFs. Random-python-scripts (November 2022) has a bleak-client BLE experiment, a Blynk config, an OpenCV camera script, a Notepad clone, a Fibonacci script, and a Design and Analysis of Algorithms course PDF.
None of these repos have a story. None of them are shaped like a project. They exist because I know, from being twenty years old and losing an SSD, that I cannot predict which file matters in three years. The one that ends up mattering is always the one I did not think to save. So I saved everything.
The cost of a private GitHub repo full of 30 MB of miscellaneous files is essentially zero. The cost of losing the one file that turns out to matter is high enough that I have felt it. That is the whole trade.
The 2025 dumps have a pattern
Between April and September 2025 I made a burst of them: obsidian, linkedindump, drawing, dumpdashdad, gyb-backup, drawings. Then obsidiandata in December 2025 (which is the one I actually use, and it has real vault content in it now). Then in June 2026 another burst: adventofcode, CapCut-Drafts (284 MB of CapCut project folders with names like 0303, 0511, 0717-copy), Sound-Recordings, OneNote-Notebooks, Recordings.
The pattern is: whenever I switched laptops, or whenever I felt the fear that a folder on my desktop might vanish, I made a repo named after the folder and pushed the folder to it. gyb-backup is a placeholder for Got Your Back Gmail backups. linkedindump is my LinkedIn about text in description.md, my Experience.md, my education.md, and a latex.md where I keep my resume LaTeX source. None of these are code. All of them are things I would be annoyed to lose.
The other thing the 2025 dumps taught me is that the same content deserves several backups in different shapes. The Obsidian vault is on iCloud, on my laptop, in obsidiandata, and in an occasional zip. The resume is in linkedindump, in a separate resume repo, in a Google Doc, and in an Overleaf project. Redundancy is the whole product.
The rest of the class
The rest of them, briefly. Dump (Feb 2022) is my first junk drawer: a battery report HTML, a KAU.HTM file with something I do not remember, a sleep study report, WiFi diagnostic dumps, and two BIOS log files. Random-c-scripts (Dec 2022) is four .c and .cpp files from a college course. Cool-websites (July 2022) is a README with one link in it, google.com/calendar/about, which I put there because the page had a nice hero animation I wanted to remember. fonts (Dec 2022, last touched June 2026) is a folder of .ttf and .otf files I use when a project needs Gotham or Cubano and I do not want to hunt again. drawing has 41 timelapse mp4s of me drawing on an iPad. dumpdashdad and gyb-backup have nothing in them but a .gitattributes. transfer-repo is genuinely empty. anomock is a small backend+frontend anomaly-mock generator I spun up in May 2026 for a quick test-data need and never came back to. Projects-list is a forked template scaffold I once repurposed into a README of one-liners describing repos, my earliest attempt at what eventually became this portfolio. PyTube-demo is a short pytube script that picks a progressive MP4 stream by resolution and downloads it, kept because the pytube API keeps breaking and having a known-good version is useful. Multimedia-Augmentation-using-Noise-GUI is a PyQt5 tool from an undergrad multimedia course that added gaussian, salt-pepper and other noise types to images and video through a GUI. Blockchain-project is a college blockchain-course submission, a JS frontend and Node server against a Solidity contract, that I never redeployed after the class ganache instance died.
Why I refuse to delete
The clean-folder impulse says: get rid of the ones with no README. It is not organised. It looks bad. What if a recruiter scrolls down.
The junk-drawer impulse says: the useful bit surfaces two years later, and I do not know which repo it is in yet. If I delete python-code-dump today, then in 2028 I will remember that I once wrote a small experiment with from itertools import permutations and I will not be able to find it.
The tie-breaker is that recruiters do not scroll that far. And if they do, they see that I keep receipts.