Real AI agent examples: what people actually got done
Every feat comes with its receipt. Real results with AI agents, each with its source and an evidence grade. Updated daily.
280 feats · +280 this week · 53% shown
Search feats: An agent, a job, an industryA live blog aggregator pulling posts from Dev.to, AWS Community Builders, Bravo LT and RedMonk, built and deployed in a single day.
Blog aggregator in one day
Claude Code built and deployed a Vue 3 blog aggregation site from a year-old requirements document, rewritten as a focused prompt.
The tool flagged abnormal values and gave general health insights, built in about five minutes from a single prompt.
Blood test PDFs to dashboard
Claude took several unstructured blood test PDFs exported from MyChart and built an interactive browser tool to explore the results over time.
Ethan Mollick received a finished Gatsby lesson plan spreadsheet for high schoolers with no obvious errors, after delegating and walking away.
Gatsby lesson plan spreadsheet
Claude with Computer Use downloaded The Great Gatsby, researched lesson plans and standards, and filled a spreadsheet on its own.
A live MVP website for the podcast and newsletter went up after about 10 hours of struggling to load it on a custom URL.
Podcast website built with Replit
The author used Replit to generate a clickable website prototype for their podcast and newsletter, then finished the MVP with help from Val Town.
Within one night Devin produced a working unanswered-emails interface using the Gmail API, after the author wrote the authentication code themselves.
Unanswered email tracker with Devin
Devin wrote a Gmail function to find emails needing a reply, built a front end, tested it locally, and deployed it to a public link.
The app went live on a custom domain, dashdiary.co.uk, for a total build cost of $35.82 plus a few dollars for the domain.
Road safety reporting web app
Replit Agent built a full web app where the public reports bad driving, including a Postgres database, image uploads, Firebase login, and a landing page.
It created the text file as asked, but after clicking a result it took a screenshot too early and hallucinated the answer.
Restaurant search with computer use
Claude's computer-use agent searched Google for vegetarian restaurants near Bond Street and saved results to a notepad file on a remote machine.
Claude opened MakeMyTrip, cleared a sign-in popup after a second prompt, and began filling the search fields; no finished booking is shown.
Flight booking with computer use
Claude's computer use, running in a Docker container, opened Firefox and tried to book a New Delhi to Mumbai flight on MakeMyTrip.
The author shipped a working summariser targeting a cut from 2-3 weekly hours of newsletter reading to 30 minutes.
Newsletter summariser app
Replit Agent built a Gmail newsletter summariser for a low-coder, with batch summaries, deduplication, archiving and export, over three days.
Operator returned the requested list of courses, fees and descriptions without further input, saving the author a few minutes of manual admin work.
Compiling a course list
OpenAI's Operator browsed the author's online course site and compiled a list of courses with fees and descriptions while they got coffee.
Operator handled tour search and lesson plan but was slow; adding three grocery items took 15 minutes and an Uber price check took five.
Hands-on Operator test
Casey Newton tested OpenAI's Operator on walking tours, a Gatsby lesson plan, Instacart groceries, and an Uber price check.
The draft was usable but weaker than accepted NSF summaries; thinking mode helped, though it began hallucinating and misreading the sources.
NSF project summary drafting
The author used Cursor in Agent mode with Claude 3.7 Sonnet and two rules files to draft an NSF Project Summary from two papers.
Small scoped fixes took about five minutes to test and merge versus roughly an hour by hand; the full-stack task failed and was rewritten.
One month testing Devin
Devin was connected to the RunLLM GitHub repo and given small bug fixes and one larger full-stack task over a month.
Operator collected data for about 25 of 100 schools before the author stopped it; o1 then built a working chart dashboard from that data.
School financials from ProPublica
Operator searched ProPublica Nonprofit Explorer for independent schools and logged their financial data into a Google Sheet; o1 then built a dashboard.
Both tasks failed. The concert search returned three wrong concerts after 33 minutes, and the Anki competitor list had incorrect data.
Operator concert and product research
Operator, OpenAI's browser agent, was tried on finding Berlin classical concerts and researching Anki competitors, using prompts and screen recordings.
After 2-3 review rounds the PR was production-ready and merged. The task took about 1 hour versus roughly 15 minutes coding it alone.
Third-party token API endpoint
Devin AI planned and built a backend API endpoint for generating third-party user access tokens, then opened a GitHub pull request.
Operator completed the restaurant search and booked the reservation, but a keto meal plan and a quilt pattern search came back weak.
Operator concert restaurant booking
Operator, tested by a member of Allie Miller's team, found a gluten-free restaurant near a Minneapolis concert venue and made a reservation.
Cursor drafted an article on their reading trends in their style, citing Kindle highlights from the vault. The author edited it by hand.
Blog drafts from Obsidian notes
The author set Cursor to imitate their note.com writing style and read their Obsidian vault, then had it draft blog articles quoting past reading highlights.
Get five like these every week
The week's strongest feats, each with its receipt, plus a short roundup of AI news. Free.
You're on the list.
A free weekly newsletter. See what's in it
No feats match this filter. Show all.
Proof marks:Shown artifact linked·Numbers a specific number stated·Claim claim only