A random mix from the whole library. Shuffle for a new set.
A statistician had argued that class balancing helps mini batch gradient descent. He had the AI Astra design and run a simulation to test the claim, and the results partly corrected him: oversampling sped convergence but hurt probability accuracy, while importance weighted oversampling kept most of the speed and the accuracy. The code was cross checked with AI, and he posted the corrected conclusion as an answer.
He takes notes on both a Boox and a Samsung device and the two formats did not mix. He built an Android app with AI that converts both formats back and forth, so he can move his notes between them.
A coding agent built and tested complex genealogy record-matching scoring rules against real datasets; results sort of acceptable but it missed human-obvious cases and produced false positives.
A Magic player built a collection manager with AI and needed every card sorted by what it does in a deck, not by the words on it. The AI ran his whole collection, 2,527 different cards, through a classifier that gave each card a fit score for 61 roles like ramp and card advantage. It took minutes and cost about 22 cents, and the AI wrote the method into a reusable skill for next time.
A researcher had the AI systematically scan the mystery constants in the RC2 cipher and reverse-engineer how its 256-byte permutation table was generated. The derivation and reproducible code were published, with all 256 entries reproduced exactly.
He had the AI configure his smart TV through its debug connection to use a filtered DNS service. Ads on the TV stopped, which he jokes is either a great morning or the start of a robot uprising.
The AI configured his arr stack plus NFS mounts, share templates and a VPN init-container in Kubernetes; the cluster is now easy to manage and debug.
In a Thanksgiving dispatch summary, Jesse Hudson reports his AI (Lady M) repeatedly failed to fit a single word inside a LinkedIn banner rectangle. His AI, Siren, then produced the banner correctly on the first try.
He built Tidehook with GPT-6 Astra and generated art. Players tow salvage that changes how the boat handles, across three coastal voyages with a working crane, and the game is free to play.
Author Matt Deitke fed an entire book to the AI and had it generate a chapter-by-chapter podcast website, with subagents handling chapters in parallel. The output was a finished multi-episode podcast series derived from the book.
Every idea is a real post from someone who tried it, summarized in our words, with a link to the original. App names are swapped for plain words.