How an App Actually Gets Made

People ask how the work gets done here, so here is one product from the first conversation to the moment it went live. Nothing skipped, including the parts that went wrong.

1. It starts with a person, not a market

Every product here begins the same way. Somebody I know asks for something, or I watch them fight with something and see what would fix it. My wife and co-founder, Jessica, needed one thing: make a video once, and have it land everywhere she posts.

That part isn't strategy. That's just paying attention to the people around you.

2. Research decides whether it's a product

This is the step people skip, and it's the one I'm actually trained for. Before anything gets built I go and look: what already exists, what the incumbents charge, how they make their money, and — more useful than anything else — what their customers complain about in the one and two star reviews. The bad reviews are where the gap lives.

That fork matters. A product that doesn't survive the research isn't dead, it's personal. I'll build it customized for the one person who needed it, and it never goes in the store. Only the ones that survive get productized.

3. Different, not bigger

The scan came back the same on all of them: Buffer, Later, Hootsuite, Sprout Social, Metricool, Publer. They are good at what they do, and what they do is scheduling. They are calendars.

Jessica wasn't asking for a calendar. The video was already made. It was already on her phone. She wanted it on the other four platforms. That's the entire job.

So Blastpack starts from the video already on your phone and sends it everywhere in one tap. It runs as many accounts and brands as you have, colour-coded so you never post to the wrong one. One caption, or a different one per platform. And it tells you exactly where every post landed.

The goal was never to out-feature anybody. It was to build the thing she actually asked for.

4. The design comes before the code

The colours weren't picked from a palette. Jessica had posted a video with lettering in it, and the scheme came straight off her letters — teal, salmon, cream, deep navy. Her colours, on her app.

The icon is an illustrated jetpack drawn for this product. Every icon we make is a scene of a real object with its own colour world, never a flat symbol on a shared background. That rule is why the shelf looks like one company made it.

5. The build, and what broke

Three things went wrong, and they're the most useful part of this post.

Registration failed for three days. Meta kept returning "Unexpected error" on developer sign-up. Meta wasn't broken. An old, unconfirmed email address was sitting in my Accounts Centre, quietly blocking the confirmation step. Reading the account instead of the error message is what found it.

Four permissions were never switched on. The first connection attempt failed on invalid scopes. Here's the part worth knowing: Meta shows that warning only to developers. A customer would have seen nothing — connected fine, then simply never posted, with no error to explain why.

A cookie was too big. After connecting, the app reported it wasn't connected. Both were true. Every Page and its token were being written into one browser cookie, and with twelve Pages that came to about five and a half thousand bytes. Browsers discard any cookie over four thousand and ninety six, silently. It's three hundred and forty two bytes now.

Then the first live test post went to the wrong Page, because the app was using whichever Page the API returned first. So we built a picker. Now it asks.

6. Live, or it doesn't count

None of that gets called finished until it's been watched working in the real world. Not a demo, not a staging environment — the actual product, posting to an actual account.

That standard costs time. It's also the only reason I'll put my name on it.

About how it's made

I get two reactions to work made this way. Either it can't really be done with an agent, or it must have been taken from somebody else. Neither is true here, and both are checkable. The icon was drawn for this product. The music in that film was written from first principles, no samples. The code was written for this application and nothing else. The research was done against real competitors before a line was written.

I'm not a programmer. I'm a market research analyst who came up through creative agencies, so I have an eye, and I can architect and process-map a business. I design the product, I decide what good looks like, and I say when it ships.

That's the division of labour. I'd rather show it than argue about it.