Posts

An Australian app build in AUD: GST, ABN and the three decisions to settle on day one

You have an app idea, a budget that does not look like an Australian agency quote, and a studio in Seoul that says five days. The first thing that usually stops the project is not code. It is the question your accountant asks: "Will there be GST on this invoice?" Here is the order we recommend to Australian founders, and why it saves a week. GST and the overseas supplier A company in Seoul is not registered for Australian GST, so it cannot charge GST on its invoice. If you are registered for GST and you buy services from an offshore supplier, you generally account for the GST yourself on your business activity statement under the reverse-charge rules for inbound intangible supplies. Practically, that means the invoice needs both parties' names and addresses, a clear description of the service, the amount and currency, the date, and a line stating that GST is accounted for by the recipient. Ask your accountant to confirm the exact wording and the threshold that applies...

A white-label build partner in one person: adding apps to an agency without hiring

A design studio I work with gets the same request about once a month: a client wants an app. Not a website — an app with logins, payments, and a store listing. The studio can design it beautifully and cannot build it, and hiring two developers to sit idle between projects is not an option. That is the situation white-label development exists for. I have been on both sides of it, as the person quoting and as the person a client never hears about, so here is how it actually works when the partner is one person with an AI pipeline rather than an agency with a sales team. What white-label means in practice The client relationship stays with you. Your brand is on the proposal, the invoices, and the meetings. I build the thing, appear on calls as part of your team when that helps, and stay invisible when it does not. The deliverable arrives in your client's accounts, with documentation, so the handover is yours to make. The parts that make this work or fail are not technical: One n...

A five-day build is not the impressive part. The handover is.

Every small studio eventually gets the same request: a client wants a working product in two weeks, and the studio has designers, a project manager, and no one who wants to spend three months on backend work. There are two common answers. Hire for the project and let the hire go afterwards. Or say yes, subcontract to the first freelancer who replies, and spend the next six weeks managing something nobody can see. Both answers cost more than the invoice suggests. A third answer is duller: buy the build from a partner who only does builds, agree the scope and the price in writing, and take delivery in a form you can maintain. What "five days" actually means Five days is enough when the scope is small and honest: one form that really submits, sign-in with two roles, one integration with a system the client already pays for, an admin screen for ten cases rather than a hundred. It is not enough when "the app" turns out to be three products. Saying that before the qu...

What I decide in the first hour of an AI-built app, and what I never let the model touch

Most of the work I do now starts with someone else's app. It runs on their laptop, it demos well, and it has been sitting at "almost done" for three weeks. So before I ever open an editor on a client build, there are a few decisions I make in the first hour, and a few things I never hand to the model. What I decide before I open the editor Who is the first person allowed in, and how do they prove who they are? What is the smallest payment that has to succeed on day one? And where does the data live if the machine dies? Those three answers decide the shape of the build. If a client can't answer them yet, that conversation is the first day of work, not a formality before it. What I give the model, and what I keep I hand over boilerplate, forms, tables, API wrappers, migration scripts — the boring seventy percent. It is genuinely fast there, and charging a client for typing it by hand would be dishonest. What I keep is the wiring between services: auth domains, redi...

Ce que je trouve dans les vingt premières minutes d'une app IA presque finie

Quelqu'un m'envoie un dépôt avec ce message : « c'est fait à 90 %, il ne reste qu'à déployer ». Avant, je commençais par la liste des fonctionnalités. J'ai arrêté, parce que la liste des fonctionnalités, c'est justement la partie qui marche dans la démo. Ce qui décide si cette app peut encaisser une vraie carte, d'un vrai utilisateur, sans personne derrière l'écran, se trouve toujours aux six mêmes endroits. Voici la passe de vingt minutes, dans l'ordre, et ce que chaque constat coûte en jours. 1. La clé qui se trouve dans le bundle du navigateur Ouvrir la page déployée, afficher la source, chercher sk- , service_role , AIza . La moitié des apps construites avec Cursor ou Lovable que je reçois livrent une clé d'API au client, parce que la clé était pratique dans le fichier où l'appel a été écrit et que le framework intègre volontiers tout ce qui commence par VITE_ ou NEXT_PUBLIC_ . Ce n'est pas « risqué en théorie ». N'importe qui...

Was ich in den ersten zwanzig Minuten einer fast fertigen KI-App finde

Jemand schickt mir ein Repository mit dem Satz „zu 90 % fertig, ich muss es nur noch deployen". Früher habe ich mit der Funktionsliste angefangen. Damit habe ich aufgehört, denn die Funktionsliste ist genau der Teil, der in der Demo funktioniert. Die Dinge, die entscheiden, ob diese App eine echte Karte von einem echten Nutzer annehmen kann, ohne dass jemand daneben sitzt, liegen immer an denselben sechs Stellen. Hier ist der Zwanzig-Minuten-Durchgang, in dieser Reihenfolge, und was jeder Fund an Tagen kostet. 1. Der Schlüssel im Browser-Bundle Die deployte Seite öffnen, Quelltext ansehen, nach sk- , service_role , AIza suchen. Die Hälfte der Apps, die mit Cursor oder Lovable entstehen und bei mir landen, liefert einen API-Schlüssel an den Client aus — weil der Schlüssel in der Datei bequem war, in der der Aufruf stand, und das Framework alles mit dem Präfix VITE_ oder NEXT_PUBLIC_ brav einbettet. Das ist nicht „theoretisch unsicher". Jeder kann ihn lesen, damit Geld au...

What I find in the first twenty minutes of an "almost finished" AI-built app

Somebody sends me a repo with the message "it's 90% done, I just need it deployed". I used to start by reading the feature list. I stopped, because the feature list is the part that works in the demo. The things that decide whether this app can take a real card, from a real user, without a human watching, are always in the same six places. Here is the twenty-minute pass, in order, and what each finding costs in days. 1. The key that is in the browser bundle Open the deployed page, press view-source, search for sk- , service_role , AIza . Half the Cursor and Lovable apps I see ship an API key to the client, because the key was convenient in the file where the call was written and the framework happily inlines anything prefixed VITE_ or NEXT_PUBLIC_ . That is not "insecure in theory". Anyone can read it, spend on it, and the first evidence is usually an invoice or a quota mail. The fix is a server route and a rotated key — half a day. The reason it costs half...