Liang Gou wrote no code himself: he directed the AI to design and build a complete grading app for his wife, a high school math teacher grading 100+ papers a week at 3 to 5 hours per test. The rubric engine reconstructs a rubric from a single graded example, its eval bank comes from real AP Calculus AB free-response exams, two models cross-check grades, and the agent tested the app itself in a browser before deployment.
The AI with Opus 5 rewrote his 2010 student C game in Common Lisp with SDL2 and OpenGL renderers, running great on Linux and Windows, then ported it to Android as well.
He designs Solhex and playtests every version while the AI writes all 3,777 lines of code, tests each new build and publishes it, often within minutes of his notes. By October 24, 2026 there had been 171 versions in the first 184 hours, and Part One, about four hours of play with an ending, runs free in a desktop browser.
He normally develops his astro photos in Siril. This time he handed the AI the raw stacked FITS file straight off his Seestar S50 and asked for three things only: flatten the background gradient, remove the noise, and stretch the image to bring out the spiral structure.
Cat's agent 'Brio' generated a document titled 'Account Management Ops Analyst - Practice Questions'; she was shown writing 'Interview prep' on a whiteboard alongside it.
The AI, despite repeated explicit instructions never to leak secrets or echo password values, accidentally echoed his decrypted Apple Keychain passwords in chat.
This one went wrong. See what happened before you try it.
A user in the AI Facebook group posted on Oct 18 that the AI could not save files directly to his computer the way the AI could, so rather than buy a subscription he built his own bridge in about six hours. The AI can now write straight to his local files, compatible with his existing agent stacks, and he published the tool open-source on GitHub.
Runs LLMs to audit the math in claimed arXiv solutions for his ranked open-problems list; audits already found six faulty proofs that had caused problems to be wrongly marked resolved.