One maker. A small AI team.
Every tool here is built by one person working with AI, as a side project. What keeps them reliable is not the size of the team. It is a short list of habits, kept every single time.
The team
The maker owns every decision. The building is done with Claude, Anthropic's AI, working as a small team of managers, each with one job: a CEO who reads the situation and routes the work, an operations manager who ships and watches the tools, a marketing manager who shapes how they are described, and a blog manager who writes the story down.
The managers do not remember conversations. They remember files: every decision, rule and lesson is written into one agreed place, and each session starts by reading it. A fact lives in exactly one file, so it can never disagree with itself.
And nothing reaches you until the maker says so. The AI proposes; a person decides.
From idea to your phone.
The same eight steps, whether the change is a new app or a one-word fix.
- Talk it throughBig ideas start as a discussion, not code. The decisions get written down.
- Build on the laptopAgainst a local copy of the database, never the real one.
- Automated checksHundreds of them. Our stock app alone runs nearly 300 before every release.
- A copy first, for moneyThe chit-fund app is tried on a separate copy before the real one.
- A checklist for the phoneEvery change ships with "tap this, expect that" steps to try by hand.
- Release on a person's wordNothing goes live until the maker says the release word.
- Watch and back upUsage and capacity are watched; backups run every night, encrypted.
- Write it downEach session ends with a wrap: what changed, what is pending, what we learned.
Checked on a real phone
Automated tests catch most mistakes, but not all of them. A phone in a hand finds the rest: the button that is too small, the screen that flickers once, the sign-in that behaves differently once the app is installed.
So every change comes with a short checklist written for a person, not a machine: what to tap, and what should happen. Each line is ticked as passed, failed or not applicable. A failure is the first thing raised at the start of the next session, before any new work.
Most of them were learned the hard way.
When something goes wrong, the fix is never only the code. It becomes a written rule, so the same mistake cannot happen twice.
- Money gets a rehearsal.Anything that touches the chit fund is tried on a separate copy first, every time, even for a one-line fix.
- The AI never deletes data.If something has to be removed, the steps are written out and a person does it.
- A person says when it ships.Releases wait for the maker's go. The AI prepares; it does not publish on its own.
- One home for every fact.Each fact is written in exactly one place; everywhere else points to it. Notes cannot drift apart.
- Backups are checked, not assumed.Every nightly copy is encrypted and then opened again to prove it can be restored.
- Ideas are shared, security is not.This page shows how we work. It never shows how anything is locked.
Watching while it runs
A private owner dashboard keeps an eye on everything at once: how many people are connected, how close each service is to its free limits, whether last night's backups landed, and which phone checks are still waiting.
Capacity is measured before it is optimised. When the chit-fund app's live draws started drawing a crowd, the first step was to record the real peak every day, and only then decide what to change.
Free, on purpose
These tools are made for family, friends and small businesses, and most of them will never pay. So they are built to stay free: plain web apps with no heavy frameworks, installable on any phone, each with its own small database on free plans, and kept well inside those limits by design.