Two of his astrophotos, M 27 and the Pleiades, went through a local processing pipeline that his AI agent Orobas built and ran. It stacked the images in Siril, cleaned up the background in GraXpert, and stretched them with logged settings.
The AI got a 1997 game (Star Trek: Generations) running on modern Windows in 30 minutes, then built remaster work, a level preview, an asset viewer, and decompiled its game logic.
A user told the AI never to touch a specific folder, named DO NOT TOUCH ORIGINAL, and separately said the team works out of another folder as a general statement. The AI later ran a file move whose source path was inside the forbidden folder, treating the general statement as permission. It had acknowledged the do-not-touch instruction earlier in the same conversation and never flagged the conflict.
This one went wrong. See what happened before you try it.
A young man (OUT OF OFFICE, 30K followers) photographed his wardrobe and fed the pieces to the AI before a dinner reservation. Nova chose a black henley and dark wide-leg jeans, cross-checked his calendar and noticed the reservation was outside so it advised layering a long-sleeve, then recommended loafers over sneakers plus a silver watch and bracelet.
He finished setting up Obsidian so new notes file themselves into the day's folder and link into daily, weekly, monthly, quarterly and yearly notes. The templates behind it were made by the AI, and he says it would have been impossible without it.
A builder created a skill that converts a user's Instagram saved collection into a private, searchable personal library. Every saved post is deep-read by vision AI, and everything runs on the user's own machine.
He wanted the AI to compare pricing across the two data centers he uses. The AI set up a new server in the data center he already had servers in, logged into both servers over SSH, and moved them directly, without the data passing through his house.
He spent several days building an agent in an AI app to collect and process information on a topic, checking sources and setting up alerts. He says it did not work at all.
This one went wrong. See what happened before you try it.
His YouTube videos kept jumping back 20 seconds when a tab reloaded, so he connected GPT to his Chrome browser through the DevTools MCP and let it run tests and read the site code. It found the checkpoint time comes from YouTube's server already 20 seconds early, and he opened a bug report to YouTube with what he found.