
Aug 28, 2026 Skrrt, Full Time This month I finished my year at Alien. From now on I work on Skrrt full time. Skrrt started in January as a side project next to my job. It has grown into a product studio for apps and AI. We build products for clients and run a few of our own. Why now Anyone can build a demo now. An AI assistant will get you a working screen in an afternoon. What it won't do is decide what goes into the first version, deal with the edge cases real users find, and get the thing launched and paid for. That part is what I've done for seven years. I shipped wallets and DeFi products, led a Telegram game that grew to 140k users, and at Alien I built a miniapp platform from scratch and an AI pipeline that reviews around 3,000 user videos a day. The tools changed every year. Getting a product in front of real users stayed the hard part. What Skrrt does Four kinds of work, plus websites: Launch a new product. A web app, mobile app or AI product, planned, built and in front of real users. Finish a stuck product. It works in the demo but not with real users. We fix what blocks it and launch it. Put AI to work in your company. An AI agent or workflow doing a job people do by hand today. Fix how your team ships. We fix the processes that are broken and set up working cycles that keep a team shipping every week, without grinding it down. Clear owners, short loops, and rules people actually like working with. Websites. Built from your design or designed with you, then launched on your domain. Every project starts with a written definition of done and one fixed price, agreed before any work starts. We don't bill by the hour: AI makes good builders several times faster, and hourly billing would make you pay for time instead of results. You own the code, the accounts and the data from day one. After launch you can keep the same people on, or take it forward with your own team. Who builds it I lead the engineering on every project. The rest of the team are people I've already shipped with: engineers and specialists I trust because I've seen their work up close, on real deadlines. I bring in the ones each project needs, so you never meet someone we're trying out for the first time on your budget. The business comes first Code is the cheap part now. The expensive mistake is building the wrong thing well. So we look at the business side of a product before and while we build it: who it's for, why they'd pay, and whether the numbers can work. If the idea holds up, we make it sharper. If it doesn't, we say so early and stop, before it drains your time and money. We'd rather put an idea in front of real people in a few weeks than spend months building it before anyone sees it. The same goes for the product itself. Every screen and feature has to trace back to something a user needs to do. If we can't say what it's for, it doesn't ship. Our own products We build products of our own too, kept separate from client work: Billateral, signed receipts for AI agent payments over x402. The waitlist is open. RAIF, a self-repairing format for structured LLM output. I wrote about it here. skrrt-sh/skills, Claude Code plugins that automate commits, pull requests and releases. If you have something to build Tell us what it is and when you want it. We'll come back with what done looks like and one fixed price. Start at skrrt.sh, or write to me on Telegram.

Aug 28, 2026

Skrrt, Full Time

This month I finished my year at Alien. From now on I work on Skrrt full time.

Skrrt started in January as a side project next to my job. It has grown into a product studio for apps and AI. We build products for clients and run a few of our own.

Why now

Anyone can build a demo now. An AI assistant will get you a working screen in an afternoon. What it won't do is decide what goes into the first version, deal with the edge cases real users find, and get the thing launched and paid for.

That part is what I've done for seven years. I shipped wallets and DeFi products, led a Telegram game that grew to 140k users, and at Alien I built a miniapp platform from scratch and an AI pipeline that reviews around 3,000 user videos a day. The tools changed every year. Getting a product in front of real users stayed the hard part.

What Skrrt does

Four kinds of work, plus websites:

Launch a new product. A web app, mobile app or AI product, planned, built and in front of real users.
Finish a stuck product. It works in the demo but not with real users. We fix what blocks it and launch it.
Put AI to work in your company. An AI agent or workflow doing a job people do by hand today.
Fix how your team ships. We fix the processes that are broken and set up working cycles that keep a team shipping every week, without grinding it down. Clear owners, short loops, and rules people actually like working with.
Websites. Built from your design or designed with you, then launched on your domain.

Every project starts with a written definition of done and one fixed price, agreed before any work starts. We don't bill by the hour: AI makes good builders several times faster, and hourly billing would make you pay for time instead of results.

You own the code, the accounts and the data from day one. After launch you can keep the same people on, or take it forward with your own team.

Who builds it

I lead the engineering on every project. The rest of the team are people I've already shipped with: engineers and specialists I trust because I've seen their work up close, on real deadlines. I bring in the ones each project needs, so you never meet someone we're trying out for the first time on your budget.

The business comes first

Code is the cheap part now. The expensive mistake is building the wrong thing well.

So we look at the business side of a product before and while we build it: who it's for, why they'd pay, and whether the numbers can work. If the idea holds up, we make it sharper. If it doesn't, we say so early and stop, before it drains your time and money. We'd rather put an idea in front of real people in a few weeks than spend months building it before anyone sees it.

The same goes for the product itself. Every screen and feature has to trace back to something a user needs to do. If we can't say what it's for, it doesn't ship.

Our own products

We build products of our own too, kept separate from client work:

Billateral, signed receipts for AI agent payments over x402. The waitlist is open.
RAIF, a self-repairing format for structured LLM output. I wrote about it here.
skrrt-sh/skills, Claude Code plugins that automate commits, pull requests and releases.

If you have something to build

Tell us what it is and when you want it. We'll come back with what done looks like and one fixed price. Start at skrrt.sh, or write to me on Telegram.