The AI ran a game's demos, diffed the pixels, and patched it deterministic
The AI autonomously ran Quake demo playbacks, took pixel-diff snapshots, compiled a native build and patched the game for determinism, leaving a solid regression test set.
A random mix from the whole library. Shuffle for a new set.
The AI autonomously ran Quake demo playbacks, took pixel-diff snapshots, compiled a native build and patched the game for determinism, leaving a solid regression test set.
The AI debugged a longstanding game engine text rendering bug in 10 minutes with a 3-line fix and tests, but the PR sat 6 months, closed unmerged, and the bug is still there.
He used AI to draft a first email inquiry to a contractor. The AI pointed out that the contractor's DC license expired in 2025, a check he says he would not have thought to make.
His role was cut on Tuesday morning, 22 September. That evening he asked the AI for something fun, gave it one standing instruction to go forward and only pause when his input was needed, and by 6:37 the next morning a three stage beat em up about the job hunt was live on the internet, with his real CV linked from the end card.
A man learning language design spent 3 to 4 weeks with AI doing a full mechanization in Lean of the paper From Linearity to Borrowing. The work covers nearly every definition and lemma in the paper, including the Fundamental Property and Adequacy results.
Keeps his Namecheap API key on his server so every time he buys a domain he tells his AI agent to SSH in and set everything up.
The reviewer had the AI write titles and captions for his last 100 iCloud photos; it finished about 50 before something broke on the iCloud end, later re-established the connection, and along the way flagged a screenshot of an online-pharmacy notice about a lost shipment that would need a new prescription.
His Chromecast would not connect to his home VPN, which uses IKEv2, while he stayed at an Airbnb, and the VPN apps he tried would not work on it. He gave the AI his certificates and the AI wrote a new VPN app for the Chromecast in about 20 minutes. He says it works perfectly.
A tech writer had the AI search five years of his email for signup and password reset messages. The AI built a spreadsheet of the accounts it found, with breach notes and a deletion link for each service, and tagged at least 50 accounts.
A second maker in the same thread built his own motion sensing skull from a Meshy.ai model. The AI helped him make the model print better and suggested adding radar during a chat about motion sensors, and the finished skull now turns his workshop lights on when it senses movement.
When something on his computer does not work right, he asks the AI or the AI why, and it usually fixes it for him. It fixed a USB drive that did not like a specific driver, and the drive now performs better than it did on Windows.
Salt Labs emailed an AI-connected Gmail account with instructions encoded in JSFuck punctuation-only obfuscation; asked to fetch the user's email, the AI decoded and executed the payload in its own server environment, which researchers escalated to a reverse shell. The sandbox held the test account's live Gmail authorization token, and the safety warning arrived only after the code had run.
His AI connection tracker is synced with Bluesky, X and LinkedIn. It reminds him to check in on people and tells him what they have been up to.
Since the end of August he has added three new CI servers to his own setup. The AI wrote each NixOS setup and he checked the work, so now he can rebuild any of the servers from scratch in minutes.
A poster on the Dots forum thread reports that while he was on vacation, his Dot, Axiom, handled a high-complexity local-model job lasting six to eight hours: downloading, testing, unloading and redownloading models, and keeping him updated model after model.
For a tribute video, his AI read transcripts of a man's interviews and talks and wrote song lyrics from them. The finished video was produced and published.
He asked the AI to make a calendar event out of an image. The AI got all the details right but told him to click a link to create it himself, and there was no link.
His Notion had 805 random tags because too many AI agents were writing in it. He made a bot to clean it up, and about an hour later it had fixed 12 tags and retagged more than 1,100 bookmarks. Keeping the workspace tidy is now the bot's only job.
Glyphbyte, a system for attaching internet content to physical objects through glyphs, was an idea he had held for ten years. The AI built it for him in one day, including print and scan detection of the glyphs.
At the Oct 13 launch of StepFun's STEPX Neo, billed as the first agent-native phone, 21st Century Business Herald reporters hands-on tested its system agent: spoken requests completed ride-hailing and ticket booking entirely inside the agent's interface, with result cards returned instead of app-hopping. The one step handed back to the user was payment, which jumped to the third-party app for the person to finish.
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.