Home › A daily AI app factory that stops and waits for approval

A daily AI app factory that stops and waits for approval

Affiliate disclosure: Mulkern AI Systems is an independent affiliate partner of Easy (easy.one), a product of Smart Apps Limited. If you buy Easy through our links, Mike Mulkern earns a commission, and the coupon applied at checkout (MICHAEL) is his. Everything on this page describes how we run our own businesses on Easy. It is our experience, not a statement by Easy.

Every day at 09:00 UTC, a scheduled job on our Easy installation researches small browser app ideas, scores them, builds the best one, and reviews it. Then it stops. Nothing goes live until Mike approves it. This page covers how the mini-app factory runs, the gate that decides what gets built at all, and the timeout bug that taught us to keep the launcher short.

What the app factory does each day

The pipeline runs in stages: research and scoring, a product-class gate, the build, and a review before anything is listed. When the review passes, the coordinator sends Mike a notification with the app name, one plain sentence about what it does and who buys it, its score, the review result, and a local preview he can open. The message ends with "Approve to publish, or say what to change."

A built, reviewed app waiting for Mike counts as a complete run. Deploying, listing for sale, and syncing it to our other sites only happen after his approval is in hand.

The product-class gate

The gate runs before anything is built, so it decides which ideas are allowed to become apps. In September 2026 Mike ruled out apps and tools that carry legal responsibilities. Ideas that fail that test are reported with the reason and are never built.

Scores rank the candidates that pass. Mike's current instruction, from 30 October 2026, is that new ideas are found and a new mini-app is built and tested every day, and the score never decides whether something is built at all. The review step can still send a weak build back for rework.

The bug that taught us to keep the launcher short

The scheduled turn that starts the pipeline used to wait and watch the work. For four days in a row it was recorded as a timeout, dying exactly at its time limit, while the pipeline underneath succeeded and delivered a reviewed app each time.

The fix was to make the launcher do one thing: create the job, confirm it was accepted, and end its turn. Easy jobs run under their own coordinator and report back on their own, so the launcher never needed to wait. A launch that finishes in under a minute is now the target.

How this maps to Easy

Easy calls a coordinated piece of multi-step work a job. A job is planned, split across agents, and run to completion while you carry on with other work. A scheduled task starts it every day, and the approval step is a notification that waits for a person.

If you build something like this, put the approval rule in the brief that the job carries, not only in the launcher. The coordinator is the one doing the work, so it is the one that needs the rule.

In the Easy documentation, see Jobs and Your first job.

Questions

More from how we run on Easy

• Scheduled AI agents with approval gates: the rule we run under
• An AI chief of staff that writes the daily report for two companies
• Running two businesses on one AI installation
• See every workflow we run