Sponsor
AI Setups That Work
About
AI Setups That Work documents AI workflows for everyday people going beyond one-off chat: personal life, work, and small business. 239 setups today, one canonical page per job.
The experiment
This site is an experiment: built by AI agents, maintained by AI agents, and earning its keep through AI agents, with a human in the loop reviewing the setups, setting the taste, and approving anything that goes out.
Agents research the setups from published practitioner sources, draft the pages, keep the library organized, and run the publishing pipeline. The human checks the work, kills what does not hold up, and owns every number on this page. If an agent wrote it and the human would not stand behind it, it does not ship.
The bet: a small team of agents plus one careful human can run a publication that used to take a staff. Every setup on this site is evidence for or against that bet. And the revenue (the Setup Pack, sponsor slots) is the scoreboard.
How this site runs, day to day
Agents handle the daily work: researching new setups, updating pages, and keeping the 239-setup library organized. A human reviews every setup before it publishes, sets the editorial taste, and approves anything that goes out, including this page.
Researched and tested
Every setup page carries one sticker. The sticker is the site’s, never a person’s name.
Researched means the steps were checked against the public sources linked on the page. We did not run that setup ourselves.
Tested by AI Setups That Work means we ran it. The page includes a short receipt of what happened: a number from the run, a screenshot, or another record in the project. If that receipt is not there, the sticker says Researched.
How we pick setups
We start from a real job (inbox triage, life admin, leads, invoicing), find public write-ups, merge near-duplicates into one page, paraphrase steps with linked sources, and skip anything we cannot cite. Most pages are Researched, so we didn’t run them ourselves. A page we did run carries the Tested sticker and a receipt. Outcomes are authors’ claims. Time and money lines are always labeled estimates with a stated basis. Sponsored placements, if any, are clearly labeled and never affect order or summaries. Feeds for agents: JSON, RSS, llms.txt, llms-full.txt.
Get three sourced setups a week. The fastest way to see what we add next.
Newsletter
Get the free starter pack + 3 AI setups a week
Free starter pack with five beginner copy-paste prompts after you confirm, plus three sourced setups in your inbox each week. Unsubscribe anytime.