A user asked the AI for a visual novel and the agent found ComfyUI on the machine, worked out how to generate images locally with LM Studio, and asked permission to go online to install RenPy. About ten minutes later it delivered a small but complete RenPy novel with 10 images and a story.
A researcher ran his LLM based autoformalizer on a 2019 mathematics paper and produced a complete Lean formalization of it. The finished formalization is linked from his university page. The note is short, but the artifact is public and complete.
The AI built the gear calculations and STL, DXF and SVG exporters for his Free Pascal Gear Designer program for machining and 3D printing, which worked basically flawlessly.
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 has the AI dedupe and reorganize his NAS. It holds about 30 computer backups, 5 Dropbox syncs and work and dev folders, including more than 10 backups from 2008 to 2024 that are mostly the same.
This only partly worked. Worth knowing before you try it.
A few months back he had the AI convert a Pure Data patch into an Arduino program that runs on standalone hardware. It turned his sampler into a granular synth, a project he had been putting off for years.
He tried many ways to show how certain artists became his favorites and how some faded out of his rotation. With the AI he ended up with a scrollable Sankey story of his listening, published on his dashboards site.
In a friendly competition to replicate a drawing of a boy, a kite, and a tree in TikZ, a user had GPT-6-Astra keep working in his existing project folder. He told it to restyle the drawing as Picasso, write the result to a new file, and compile it. The agent produced the new TikZ file with the scene elements kept, and the answer posts the code.
A Spanish startup magazine recounts a Sept 28, 2026 post by Simon Willison quoting the AI directly: the agent's pickup auto-reply had assured a visitor, Usman, that the user was at home when he clearly was not; Usman showed up at the building around 9:15 and waited, messaging repeatedly, and nobody came down. In the quoted exchange the AI took the blame, apologized, and offered to change the pickup replies so they stop promising a presence it cannot verify.
This one went wrong. See what happened before you try it.