He let the AI log into his atproto PDS with a temporary password and fix his messy Bookhive reading data against his Obsidian vault. Ten minutes later, everything was in sync.
Handed code, data and a writeup to Astra and Opus 5.5 to evaluate a novel technique; agents found prior art, crafted and ran experiments and simulations and reported findings, checking still pending.
This only partly worked. Worth knowing before you try it.
A podcaster connected his bank accounts to the AI. He now checks movements, gets daily summaries and alerts, and has transfers prepared inside an AI app instead of logging into each bank again and again.
The MBI Deep Dives author writes that the AI tracks a set of data points he cares about and reports daily, including GPU prices, the App Store ratings race between the AI, the AI and the AI, and insider buy/sell transactions in stocks he owns. He describes using roughly 200 million tokens a week on a $20/month plan.
A Purdue professor built Docable, a working tool that turns PowerPoint, Word and PDF files into accessible versions for new disability rules. The AI and the AI helped him create a codebase of about half a million lines in a few months, and his talk explains the checking method he uses so agents can validate their own work.
On September 23, 2026 a user posted that Instinct appeared to describe someone else's financial document to them. Founder Noah Shinn responded the same day that the model had fabricated a proper noun — a hallucination, not a data breach — and that the team built a hallucination detection layer within 48 hours.
This one went wrong. See what happened before you try it.
An AI user runs the agent on a headless home server with his DoorDash and email credentials stored in 1Password. The AI logs into DoorDash with a skill it built itself and chats with support reps, and the user says it now gets refunds issued with no issue where DoorDash had turned stingy with him.
The AI redlined OEM contracts through the first three negotiation rounds, finding things his lawyer missed, with the lawyer only doing the final check and money saved.
He was about to pay about $700 for a Mac mini at Micro Center on his usual card, which gives 1.5% back. The AI reminded him that his US Bank card was paying 5% back at electronics stores that quarter, so he got $35 back instead of $10.50. The AI also logs into his card portal and turns on all the store offers for him, and pings him before good offers expire.
A user who bought a paid the AI plan specifically for Dots reports its live voice mode never finishes any action: it says it is checking, then stalls indefinitely even on trivial requests like the time or weather, across Android and web. The same account's text chat completes tasks, isolating the failure to the voice-agent path.
This one went wrong. See what happened before you try it.
Crawford Collins (verified, 1.1M followers), a busy new dad, personalized the AI as 'Atlas' (fox avatar) and used voice commands while caring for his infant son. Atlas generated an efficient 4-day-a-week workout plan synced to his calendar, found and ordered fragrance-free size-3 diapers (approved on screen), and when grandma was watching the baby, found a date-night sushi restaurant (Yasu Beverly Hills, Sushi Note) which he booked.
Marco H asked the AI to audit his 15,000 Instagram followers and identify which ones don't follow him back. The agent worked through the full follower list and returned the non-follow-backs.
He runs all the steel on a large construction job. The AI builds him a packet for each item from the project document database, and reads the specs and codes to work out which one takes precedence when they conflict. He says it saves him at least 15 hours a week.
A researcher used AI and the AI Fable 5.1 to search for examples, write and run verification code, develop and check the proofs, and draft the paper for a triangulation problem. The result settles the value at 5 for every n from 9 up, with coordinates and edge lists published in a repository. He checked the argument himself; the paper is not yet refereed.
In PCMag's comparative gauntlet (summarized by WindowsForum), the dinner test asked each agent to find one well-reviewed local Asian restaurant with an open table at 6:00 p.m. that evening. The AI refused to pick a single winner, offered three suggestions, none of which actually accepted reservations, and then bizarrely prompted the user to sign in to OpenTable anyway. Dots found a 6:15 p.m. slot and flagged it as the closest to the request.
This one went wrong. See what happened before you try it.
The AI can create, snapshot, delete, start and restart VMs on his garage Proxmox box and spin up spot instances for hardware tests he does not have locally.
After dry-running a personal-assistant the AI through August, its owner let it pay a real $3,731 bill that arrived by email; routines now watch his Gmail for invoices three times a day. The first live payment went through under the watch setup he had tested for a month.