Body Composition Simulator
Food/exercise tracker that shows ‘look like this in 3 months and a year if every day/week like this day/week)
Discussion: https://chatgpt.com/c/6a90e134-66e0-83ea-a308-ea4f93ee5e92 Initial MVP: https://claude.ai/chat/7ddb583e-4479-4b7b-bdb2-e9a21e9dc851
Backrooms Your House
‘backrooms your house’ - find photos on Zillow (or upload) and repeatedly have a model ‘simplify this room by 20%’. Purposefully use a not great model which is cheaper and more likely to be backsroomy
Test image gen prompts: https://chatgpt.com/c/6a90e2b9-a0a0-83ea-acb8-6bffb086c2ca Theory: https://chatgpt.com/c/6a90e9b2-6704-83ea-85b7-65b2b4f4a28e
Superdiff
I feel like I’ve been this a dozen times. Superdiff is meant to be a document comparison tool built for long, revision-heavy documents. A normal diff is great for code, but it can be miserable on Word/PDF exports where page numbers, tables of contents, and formatting churn can bury the real edits.
Superdiff aims to surface the changes a reviewer cares about:
- Additions, deletions, and rewrites in sentences and paragraphs
- Meaningful edits to clauses, requirements, or definitions
- Updates to figures or tables that change content, not just layout
At the same time, it tries to ignore the high-noise categories:
- Renumbered pages and sections
- Table of contents refreshes
- Pure formatting or style tweaks
The output is an annotated report that calls out substantive edits cleanly. It also computes a simple “change density” score, and can suggest a rough review time based on how much content moved.
I’ve written a version in PHP, C++, js, python… I usually get it to a point where I can run it on an entire work corpus, run it, then forget about it. The need seems to resurface every couple years.
I think this is like ToDo Apps, where every programmer seems to have made one. With so many samples on github, next time I’ll probably just use AI to code/do it
Semantic Diff is what I’m using now.
Auto-censor kids’ books for comedy
This project takes children stories and adds “unnecessary censorship” for comedic effect. Automatically pick a few words or phrases and replace them with black bars or symbols, as if the text contained something explicit, even when it does not.
Example:
“The princess sang to the frog happily in the forest.” becomes “The princess ██████ the frog happily in the forest.”
It is basically a small algorithmic prank: the story reads normally until a censored block appears, then your imagination does the rest. It’s adjacent to “This Week in Unnecessary Censorship” (Jimmy Kimmel) and the broader unnecessary censorship meme… although I didn’t find these until searching to see if anybody else had this app already.