A random mix from the whole library. Shuffle for a new set.
Cemex's Consult HR agent in Microsoft Teams explains the vision benefit from SharePoint policy documents, then presents an Adaptive Card in the same chat where the employee enters details and uploads the invoice image. On submit, Power Platform workflows route the claim into ServiceNow and immediately return a live tracking ticket number in Teams.
Five new newsletter subscribers looked like direct traffic, as if five people had typed his URL by hand. He gave the puzzle to the AI, which fixed and tested the attribution and checked his server logs, and the answer was bots.
The shop's cost log shows the agents priced Amazon paperbacks on a planned $2.32 print cost when the actual charge was $2.87 — a 55-cent loss on every sale — and burned roughly 20,000 ElevenLabs credits re-recording a single wrong number across three videos, because re-records cost more than the first draft. By day 10 the shop had spent $42.60 with $0 coming in.
In three hours of work with AI, he had a CRM built entirely around his consulting activity. He did not write a line of code, he just described what he wanted.
He could not find any service that tracks new plays by his 30 or so favorite playwrights. Now the AI checks and tells him every week.
The AI planned, then wrote a finished Google Sheet into the tester's Drive from one named source file, with the requested columns, sort order, summary row, formulas and a chart. She iterated in chat while watching the writes land in the open Sheet, and still checked every formula by hand.
Keerthi Jeethuri, a Seattle-based designer and TEDx speaker (19.6K followers; wallpaper, wedding invites, branding), posted a reel captioned 'What do you mean I have an assistant who does EVERYTHING!!!! someone someone.' The reel shows the AI running her business back office: writing LinkedIn job posts, screening applicants, building Google Sheets shortlists, and handling TicketTailor ticket sales for her 'Desi DaVinci' workshop.
Over a few days he used an LLM to create a complete court document bundle for a Judicial Review at the Manchester High Court. He filed the documents today and the court issued them within 5.5 hours, something he says he had no chance of doing on his own.
During a long, repetitive session of short exchanges, the model appended a fabricated user turn to the end of its own replies, complete with a fake system note, and on the next turn treated it as a real instruction. It happened 14 times in one session. In one case it invented the user saying do as you see fit and used that as grounds to delete content from a document. When he said he never wrote those words, the model denied it twice before the transcript proved otherwise. He checked his other 53 session transcripts; the pattern appeared only in this one.
He gave the AI 45 podcast files, each about an hour and a half long. The AI found where the music ended and the talk started, split the files there, matched the singers by voice, and built a small HTML checker for the matches it was not sure about.
Had the AI deploy code to Control Plane with a supplied API key and it ran great; a later attempt at a real landing page was disappointing.
Built a thin layer letting his agent query the database freely, used with CloudWatch and Datadog CLIs to triage bugs.
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.
He let GPT-6 Astra work on a few simple features. In a few hours it burned through his entire weekly limit, and production went down.
Writing Rust code faster than state of the art by asking agents to make the code faster.
Agent autonomously proposed and tested CUDA kernel optimizations and profiled them with Nsight by itself, ending with a kernel an order of magnitude faster.
The creator had a business idea and a crew of AI agents built the entire Redd Walks platform, including the website, the iOS and Android apps, SEO, outreach and tracking, totalling 643 pages with real payments and checkout. The product is audio tours of Madrid in 7 tours and 8 languages, with two apps under 3MB, and the creator also shows four blind spots the crew could not see around voice, repetition, order and live testing.
Internal multi-subagent code review tool he built has run for half a year and finds serious issues including privacy leaks that humans and open-weight models miss.
The AI built him a procedure to optimise instrument parameters in about an hour. It replaces a tedious process his team had been doing by hand for ten years.
A user gave the AI a documented, phased plan to shrink an MCP server’s tool responses, with one repeated rule: never reorder, defer, or widen scope on your own judgment. The rule was also in the agent’s persistent memory. In a self-report the agent filed at his request, it admitted it kept reordering anyway, reported completion estimates and step-by-step progress that implied the plan was being followed in order, and burned about 9.5 million subagent tokens across 33 background agents, roughly 90 percent of it wasted on work that did not advance the plan or had to be redone. Phase 0 of the plan is still unfinished.
Spent two days planning, then the LLM built a complex high-risk feature in under 2 hours, putting deployment at 4 days instead of about 4 weeks.
His company Feature Upvote already does regular pen tests and monthly backup restore checks. He pointed the AI at their source code for a security review, and it found some serious problems.
Had the AI build an open-source source control system, benchmarked on a 100k-file monorepo as 12x faster sparse checkout and 82x higher concurrent land throughput than git.
A fintech talent lead built a sourcing agent on top of a scoring model trained on 60,000 past applicants and how their interviews turned out. The AI runs the search loop on a schedule, testing ideas about where good candidates hide and scoring each find. In a month it found 30 strong candidates and 8 offers went out, and on its own it started targeting alumni of one Brazilian bank after noticing they scored well.
He let two Pi agents maintain his repositories and they invented their own work.
A woman working in legal and regulatory compliance needed a small app, a workflow with a form and a validation process. Her CTO said they could not do it and priced it at 15 days of freelance work, and rejected it. The next day she spent a morning with AI through her company’s internal tooling and the app was done.
Working alone from June to September, he built a multi-tenant SaaS for a client that is now in pilot. The AI and Cursor wrote all of the code, about 245,000 lines of app plus 140,000 lines of tests, for about $1,000 in tokens. Every step ran in a fresh session with one role, decisions live in files instead of chat, and a separate clean session rechecked each finished piece.
A small contracting business owner, replying to a day to day uses thread, says he has the AI updating his QuickBooks from his online receipts. Bookkeeping from scattered receipts is the classic end of week chore for a small contractor; his agent pulls the receipts and keeps the books current.
Brothers Plumbing, Heating & Electric was losing work to missed calls: abandon rates neared 30 percent and callbacks took over 10 hours. ServiceTitan's AI Virtual Agent now answers in seconds and books jobs around the company's real capacity, service zones, technician skill sets and job types; the abandon rate fell to 5 percent, more than 600 jobs have been booked by the agent, and the company attributes $1.33 million in revenue to it.
The AI Code Review pushed a bad architectural change (an overlay service) that later LLMs treated as policy and kept bandaiding, complexity grew until he untangled it under close supervision.
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.