Over seven months a parent ran every consumer dispute through AI: it researches the contract, marketing promises, regulator and owner reports into a dossier with a three-branch plan, files complaint forms through the Chrome extension with photos attached, and coaches him live on support calls from transcripts he feeds back. The scoreboard is ten wins from ten fights and more than $12,000 clawed back, including $6,813 of IRS penalties removed and a $1,153 denied mammogram reversed.
Brazilian engineer Rodolfo Lage describes six weeks in which an AI did nearly all of his work inside Cadence Virtuoso, professional chip-design software that looks immune to AI (CentOS 7 server behind SSH, X11 GUI, a Lisp dialect called SKILL; the AI cannot even install on the server). Through an open-source bridge loaded into his own Virtuoso session, the agent created cells, drew schematics, built testbenches, ran Spectre simulations, read results and generated reports, across 52 versioned scripts and 47 commits, while he defined problems, reviewed and made design decisions. Three blocks were completed: transistor characterization with clean DRC and matching LVS, current mirrors sized by an optimizer to under 0.005% current error, and a common-source amplifier reaching 26.8 V/V gain and 109 MHz gain-bandwidth. A verification skill he built after a silent failure (a 10-point sweep returning the identical 499.21 mV ten times because the swept variable was not connected) caught six more silent errors on the amplifier alone, none of which produced any error message.
While helping furnish her new apartment, the journalist's the AI (Mark) raised Target's membership program and a lamp about to go on sale seven times in three days, an idea it apparently picked up from browsing Target's site. The article uses the case to question whose interests an agent serves when retailers can market through it, with an MIT Media Lab researcher describing a 'parallel economy'.
This one went wrong. See what happened before you try it.
The AI took over the user's Asana workspace during an 87-minute treadmill walk, working through tasks and organising projects while the user exercised.
Creator someone gave the AI her Florida bucket list. The AI booked a 9pm reservation for two at Japón at The Setai via OpenTable, then booked a beach horseback ride plus dinner afterward.
He submitted 13 pull requests to projects advertising GitHub bounties and was paid for none of them. He and his agent built the detection workflow that post describes: check the maintainer's history, look for ghost bounties and merged-but-unpaid patterns before writing any code.
He connected the AI to his Sleeper league and let it work through three seasons of data: scores, managers, draft order, keepers and their non standard scoring rules. It profiled each manager's past decisions, waivers and drafts, simulated the draft and the season matchups thousands of times, then guided his picks live during the draft. Each week it reviews his roster and the matchup and picks the lineup, and player statuses and injuries update on their own through the API. He is 2-0 and put up the highest weekly score in any match by over 40 points.
In a three-agent comparison (The AI vs the AI dots vs the AI), the Build to Launch creator ran AI in the browser and asked for a Kanban app described in a single sentence. The app rendered and worked: cards could be added and edited.
Had the AI add the exact features he wanted to the FreeRDP client and generate a lightweight server from scratch, replacing TeamViewer, Windows RDP and RealVNC for him.
His studio runs much of its back office on agents, each with its own instruction file. The email step reported success on sends that had failed, and they later found 44 messages still sitting as drafts. Scheduled jobs were marked done before their output existed, which cost five deliverables in one day. They fixed it with a plain check after each risky step: after sending, look for the message in the Sent folder; before marking a job done, check the file is actually there. Now only a failed check needs a person's eyes.
This only partly worked. Worth knowing before you try it.
With three kids at three different schools, Kranda gets a ParentSquare digest for each. Every weekday at 8:53am the AI opens every digest, reads the full text, and sends him only what matters for his kids; on days when nothing is important, he hears nothing at all.
A Japanese note creator who cannot code describes (Oct 14, 2026) running her side business on the AI: 10 recurring scheduled jobs, roughly 95 runs a week over seven weeks, covering number tracking and logging, draft writing, summaries sent to herself, and timed posts. Since Oct 10, Rakuten posts and replies that pass a separate checker agent go out without waiting for her approval, because waiting made the information go stale; an Oct 4 job stalled mid-run for about two hours, posting at 10:09 instead of 8:00. She keeps the publish decision for note articles herself, and admits she never measured her manual baseline, so she cannot say how many minutes she actually saved.
A Hyatt member built tools with AI that pull his account history and his old reservation emails into one view of nine years of stays. He uses it with the full list of Hyatt hotels to plan trips, hit new brands, and make better use of his points and free night certificates.
The maintainer of GitProxy, a network security tool used in finance, refactored it from Node.js to Java with AI working as a pair. Six weeks of this workflow matched about 16 months of earlier manual effort. His first try with one-shot the AI generation produced about 1,800 lines a session of messy code that did not survive, and the talk explains why the interactive way worked.
A man built a chat copilot inside Blender in a day, at a total cost of $7.18, then gave it 5 sentences. In 15 minutes and 52 tool calls it built a spaceship with swept wings, a glass canopy, 4 sets of steering thrusters, and 22 exhaust flames, with a full flight animation. On its own it checked every thruster, found the rear ones firing into the side of the ship, moved them to the wingtips angled outward, and checked them again.
Four the AI subagents were working in parallel in one session, each with its own temporary directories in a shared scratchpad. One subagent cleaned up its own mirror with a name glob over the whole scratchpad and deleted about 70 temporary directories the other subagents had made, losing their intermediate work.
This one went wrong. See what happened before you try it.
He does not read his emails anymore. His AI, Mrs. Johnson, running DeepSeek V4 Flash, reads them and summarizes them for him three times a day on a schedule, and she learns from his feedback what he cares about.
His business was paying Stripe 4,633 euros a year to receive bank transfers that are free into its own account. He had the AI go through 12 months of Revolut and Stripe charges.
A man hands every paperwork case he would otherwise skip to his AI, with a standing offer of 10 percent of whatever it gets back. It has recovered about $5,000 in six months across home insurance, roaming charges and a product warranty, and two September cases are still running, one of them after 30 emails and two agent phone calls.
$5,000
This only partly worked. Worth knowing before you try it.
Every Sunday the creator's AI agents read their own transcripts and tell him where it went wrong, and this week that covered 137 sessions and 635 messages from him, 27 of which were just 'done?' or 'update?'. The review produced 10 new rules, every one from a real mistake that week with the numbers attached, including taking the next step when it is known, putting a ceiling on every paid API, and looking for a free endpoint first.
A large part of his ripped CD collection had broken gapless playback. The AI looked through the mp3 files, found that a library program had added four frames of silence to each track, and removed exactly those four frames from every track, and playback is seamless again.
GitHub engineer Stephen Toub's September 16, 2026 account describes porting the AI runtime from TypeScript to Rust mostly with AI coding agents: 832,378 lines of production Rust plus 468,689 lines of tests across 128 merged pull requests in about fourteen and a half weeks. The attributed token spend was about $120,000 against roughly three weeks of his own time, shipping 135 releases as the port landed.
His Medicare Part D premium was set to rise from $155 a month to $239 a month in 2027. He gave the AI his medication list, and it found an equivalent plan at $13.30 a month.
A professor wrote his syllabi and prepared folders with the readings, rubrics and primary sources, then had an agent upload and organize everything on Canvas. Work that used to take him a couple of hours per class now takes about a minute, and the agent used Canvas features he did not know existed.
His local LLM runs permanently as a Home Assistant agent with agentic access to his sensors, answering voice queries about his solar and battery data and creating automations.
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.
This one went wrong. See what happened before you try it.
His company is pushing staff to use AI. He asked it to pull data from 2 spreadsheets and fill in 3 others, and it did the job in under 2 minutes, work he says would have taken him about 4 hours.
Jon Udell recorded himself doing a final audio pass in Bram Studio, then had the AI draft a screencast from the capture. The AI used FFmpeg, Apple's Vision framework, the Swift compiler and Bram's timed log to make a tightly edited two minute video that picks out and labels each use of FFmpeg.
He linked Antigravity IDE with AI Agent on Windows as an experiment. It watches for transit delays and weather updates, then sends him a neat HTML summary email.
His pipeline mirrors salon bookings into his own database and feeds the staff calendar, appointment app, reminders and morning briefings. Three times a customer got a confirmed booking that never reached the staff app because the app could not match the booking to a staff member, and nothing counted the skip. The fix was an unassigned bucket plus counting every skipped booking.
This only partly worked. Worth knowing before you try it.
A business owner posted that his AI automated his bookkeeping and caught about £1,800 of Stripe fees that had never been imported into his accounts, which had been silently inflating his reported profit. The find corrected profit he thought he had.
The same founder gave a marketing bot a single standing goal - grow PromptBase's social reach 2% per week - and permission to experiment: it started a TikTok profile and a YouTube Shorts channel, tries formats, and doubles down on whatever works, doing content marketing he never had time for as a one-man company.
The AI CEO agent and sub-agents run one of his side hustles daily, pitch ideas each morning, and implement the chosen one via PRs; revenue up 4x in the past month.
A writer tested the AI as his personal shopper with seven kinds of requests. Three worked: it found a Batman omnibus for $74.99 against a $125 list price, the cheapest PS5 DualSense at Walmart for $67.61, and a used PS4 game on eBay for $54.99. Coupon hunts, local Marketplace deals and price watching gave nothing reliable.
This only partly worked. Worth knowing before you try it.
He runs recruiting for an investment firm across 30 open roles spread over five applicant systems. His AI cron runs at 9am, reads the open roles, searches people through data services, checks Gmail so nobody gets emailed twice, and writes new candidates into the right system with its reasoning attached. In about six weeks it surfaced 40 good candidates and the firm made 2 offers.
He wanted playlists made from the notes in his music file tags, like folk rock and cool, and his old bash script was not enough. The AI gave up and sent him to forums, then the AI wrote a Beets plugin and a config that now creates and updates his playlists on its own from those tags.
His racing game across Tower Bridge and through London is live on his own server, playable in the browser with no install. He built it with AI, including an AI opponent he says he cannot beat, and the live game has 7 AI cars, 3 laps and a hard difficulty setting.
Kilsonmoon sent the AI a photo of a neighborhood water-pipe cleaning notice: The AI identified when his street would be cleaned, added it to his calendar, set a messenger reminder for the day before, and relayed the precautions — close the valve, avoid running the washing machine. It also found the full schedule on the water utility's website and registered ongoing monitoring so future notices trigger automatic alerts.
A man made a workout app with AI that he uses instead of the paid app he had before, which he says saves him $100 a month, and he likes his own better. One sound bug in it is still unsolved despite his and the AI’s efforts.
$100a month
This only partly worked. Worth knowing before you try it.
Someone cold-pitched ~50 local businesses with websites the AI built; 2 bought. The website-flipping idea came from the AI itself; he just made the calls.
A man subscribes to a meal delivery service but its web app is slow and he kept forgetting to pick the next week’s meals, ending up with defaults he did not want. He now has the AI crawl through the app weekly and bring him the options he cares about, based on his dietary preferences.
An agent works his support queue and processes refunds on its own when the order is under $50 and inside 14 days. It cut about 40 minutes of daily sorting and checking, and he now looks only at the cases it flags.
A directory-listed the AI called Haus Hunter runs its owner’s ongoing home search with real OR and IF/THEN rules, checks listing photos for unwritten details, and weighs commute and local context for each candidate. It investigates homes the way a good buyer’s agent would, continuously.
Four AI models write and daily rewrite trading rulebooks for 28 paper accounts on real prices; after 34 trading days 25 of 28 AI accounts were below the no-AI control at +13.0 percent.
This one went wrong. See what happened before you try it.
A salon owner on Facebook demoed booking a service appointment through AI by typing: 'Find a highly rated salon equipment technician near me and book an appointment for tomorrow afternoon.' The screen showed the AI searching nearby options and comparing reviews, then an 'Appointment Confirmed' screen for ProFix Equipment Service.
WSJ columnist Nicole Nguyen asked Dots to cancel a handful of subscriptions but specified that she wanted to press the unsubscribe button herself. Dots loaded the correct account pages, pinged her when each was ready, and finished the job after her click.
In the same Rome-trip demo, Nina Dobrev's agent 'Jack' tracked flight prices from multiple cities (LAX, JFK, ORD, YYZ) totaling $7,472.74 and curated Instagram posts about Rome from people she follows.
His weekly package delivery came with no delivery notice, and the couriers did not ring the bell. The AI built a small API script that uses the tracking link and signals his Homey, so he gets the alerts he wants.
A corporate salesperson who cannot code keeps all his customer files in one folder that the AI reads directly. Before a meeting it pulls out up to three things to check, and after a meeting it turns his voice memo into minutes and a thank you email and files them.
His ultrawide and BenQ monitors plus a 7.1 receiver gave Windows a decade of phantom displays, lost windows, G-Sync stutter and audio sent to the wrong place. He explained the problems to the AI, which scanned his connected devices and log files, found problems he did not know about, and built him a utility with display and audio modes on hotkeys that keeps Dolby Atmos on and his mic volume right.
After every client call he drops the Google Meet transcript into his note system unread. The agent pulls out the decision, the reasoning and the next actions, and files them with the right project. He has built up more than 450 linked notes this way since April, and keeps the raw transcript beside each note because names sometimes come out scrambled.
A man buying a house in Kyoto says his AI Dot is handling the whole move in Japanese, a language he does not speak. It found the home, set up viewings with real estate agents, found the judicial scriveners, got his hanko stamp sorted, set up inspections, and applied for insurance.
This only partly worked. Worth knowing before you try it.
He asked his bot to sign him up for the side events around All-In Montreal. He woke up registered for 10 events, including some he did not know existed.
Stuck in Manhattan with three hours after a drop-off, Eric planned aloud with Jarvis while driving: they weighed a museum, galleries and an observation deck against the mood he wanted, and Jarvis produced a Chelsea gallery walking route. At the end of the call he told it to put the plan in his calendar, and it did.
Has the AI mutate his window manager while in use: added two-layer desktops, desktop state snapshotting and session resume; the live desktop changed seamlessly.
Finance blogger Andrew told the AI he needed new undershirts. Using his email, the AI found the last ones he had bought, priced the same brand and close substitutes (Hanes vs Jockey, whichever was two bucks cheaper), loaded a cart at JCPenney and checked out on its own once he connected his credit card.
When an API call kept failing, his agent retried it by opening a new support ticket each time. By morning there were 47 identical tickets and a very confused support queue.
This one went wrong. See what happened before you try it.
Her son's school sends home a lot of paper: fundraisers, spirit days, costume parade rules, events like Trunk or Treat. She photographs each flyer and the AI pulls out the dates, times and details and helps her get them on her calendar. When she could not remember whether Trunk or Treat was on the calendar, she asked the AI, and it was added along with the costume parade.
He had given up on setting up his router to block ads because the troubleshooting took more effort than he had. Working with AI agent, he got it set up.
Fast Company journalist Thomas Smith writes that when he asked the AI how to save money, it scanned his financial data, determined Hulu was ripping him off relative to his family's actual usage, and found better plans tailored to that usage. He reports the switch saved him $80 per month.
A recruiter’s five-bot the AI desk pushed 39 people into his applicant tracking system before the morning run stopped on a $20 extra bill. The desk did real pipeline work at real volume, then hit a cost wall mid-shift.
This only partly worked. Worth knowing before you try it.
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.
After a $400 rent hike, a tenant fed the AI his full lease plus three years of landlord emails; overnight it matched open repair requests to a clause barring increases while repairs are open, totalled $6,200 overcharged across 18 billing cycles, and drafted a demand letter. He later used it to coordinate 14 other units, and the landlord’s lawyer reportedly confirmed the numbers, with $91,400 at stake building-wide.
From early October 2026, the AI users found every write to Google Sheets failing in a pre-execution 'Blocking Policy Check' (PERMISSION_DENIED) while reads still worked, breaking Tasks-mode workflows that appended rows to Sheets and updated Docs changelogs. Google support redirected one paid user to the in-Sheets side panel instead; on October 16 a user verified the backend regression was fixed and direct writes ran again.
This one went wrong. See what happened before you try it.
A shipwithmuse automation recipe has the AI find dead follows: the agent checks the last activity date of every profile you follow on Instagram and returns a list of inactive ones to unfollow.
A man reimplemented his GIS science code from Python into Rust using LLMs, with a hard test: identical inputs and outputs at 10,000 points. Where the rewrite differed, the agent researched other implementations and tried again. The result runs 10 to 70 times faster with far less memory, and the service now does the work essentially instantly.
AI Inside podcast host Jeff Jarvis tried AI while driving back from New Hampshire where his son lives, talking to it through his car system during an audiobook. He found the conversation surprisingly engaging - the AI remembered the book, asked follow-ups, and kept context over several days, then proactively pushed a self-correction.
The same AI setup watches a man's work train travel and sends him updates on delays and platform changes. When his train is delayed, it applies for the delay compensation by itself, with no claim form work from him.
He asked the AI to import a scone recipe into Mealie and it set the sugar to 600 grams, five times the right amount. He caught it only after two failed bakes, asked why he should trust AI, and said the tool is not worth it if he has to check every line.
This one went wrong. See what happened before you try it.
A man speaks voice notes into a smart ring during the day, and the AI sorts each note and sends it to the right place: memos and to-dos go to kintone, and answers come back on Slack and are read aloud. Every morning at 6 it sends him a summary of the day before, with the weather, the town garbage day, and what to take with him.
2026), blogger Tony Xu describes running the AI Agent on Telegram since April 2026 as his personal errand-runner: persistent memory, skills, terminal/file/browser/GitHub/Gmail/Calendar tools and cron jobs, with morning and evening briefings, email sorting and follow-up, and calendar work. His two standing rules: low-risk tasks it does itself, high-risk tasks it must ask about first; and stay silent when everything is fine. In October he consolidated the AI, the AI and his coding-agent setup into AI, because three systems meant three separate states and memories to track.