Amin Sweiti

I'd learnt more about startups and business in 3 books than in the year I spent in "business" school, so I dropped out. Everyone says the first hurdle is starting something, so I wanted to get some failing out of the way.

I saw this KBBQ table on social media from an American company and realised no one was distributing them to Australia. That day I found their suppliers, called my 2 friends and ordered 1 unit straight from China.

The plan was simple: shoot a shit ton of content, see how many pre-orders we could get, then order in bulk. Worst case, we have a cool KBBQ table.

The table arrived absolutely destroyed. We couldn't afford another, so we cleaned it up and shot content anyway. We were $1000 and 4 full days in, posting mostly to no avail, and my friends had lost hope. I kept posting until one video BLEW THE FUCK UP.

5 million views. 10k followers. In its first week.

Orders flooded in from all over the world, $30k worth. We thought we were rich, until we realised a 20kg table that's a metre long isn't easy to fucking ship. It would cost $1000 to ship our $500 table to America. So we refunded every international order and ordered 20 tables for our Australian customers.

The next morning: YOUR SHOPIFY HAS BEEN TERMINATED. No store, no customer info, just one form to fill out. Customers got worried and did chargebacks, and by the time the shipment arrived, only 10 of the 20 were left.

But at least now we had 35k followers. We put up a story, sold the last 10 that week through DMs and rented a van to deliver them all ourselves.

In the end we made $1500. 2 months of work, $500 each. But now we have a KBBQ at home and an awesome story to go with it.

During Gachi Grill I'd been doing Harvard's CS50 online while fucking around with AI. I'd never really coded, but I've always had a math-oriented brain so I picked it up quickly. It was clear AI was going to get better, so I focused on core software engineering concepts rather than specific languages.

Gachi Grill also showed me the value of a personal brand, so I started launching software projects with the goal of making my way to San Fran.

I made my first app and released it in a week. The point was to learn how to release an app, and I could easily feature it in my focus timelapses.

Off the back of Gachi Grill, an American founder reached out to me about LiveChance, a casino platform merged with a live streaming platform. I grew up on games and streamers, and it let me work from anywhere, so I accepted the offer and moved to Southeast Asia the following month.

We needed streamers, so I built a script that scraped everyone going live in Kick's gambling section into a Google Sheet with their contact info. In a month we had 4000. About 1000 were in the right states. I reached out to all of them and did about 200 calls. 10 did test streams and 50 agreed to try the site without any initial payment.

My side was going to plan, but the site wasn't. The founder wasn't technical and relied on an iGaming dev team. The launch was meant to take a month and took 6. Then at soft launch, the first depositor accidentally found a bug that let him add money to his balance without depositing. The devs blamed the payment provider.

I got the codebase and got them on a call, and realised the founder had been getting absolutely screwed for the last year.

That day I learnt everything about the payment provider, found the issue and fixed it along with a long list of neglected bugs. We relaunched the next day.

I went from CMO to CTO, but it was too late. The founder had 1 month of runway left, no funding, and the streamers had been let down by months of delays.

While at LiveChance I had my eyes on my own project: an app that merges reading and listening, so you can switch between the two or do both at once.

Speechify, ElevenLabs and the rest disappointed me as someone who actually likes to read. They all treated TTS as a solution to reading, not an enhancement.

I started with premium cloud voices and got Inworld running. It sounded awesome, but its timestamps were unreliable at the time, and word-level highlighting is essential to my app. The numbers didn't work either. A $35 a month plan could only give you 40 hours generated.

So it had to run on the phone. No server, no cost per hour, no limit.

Kokoro had the best sound of anything with word timestamps, but it was too slow. The best numbers I could find on an iPhone were 3 to 4x realtime on the GPU, and iOS takes the GPU away when your app goes to the background, which ruins the whole point of my app. On the CPU it ran at under 2x.

It wasn't good enough. That was on an iPhone 17 Pro, so older phones would be way worse and the phone would get hot because it never stopped generating.

So with 0 experience at this level, I changed how Kokoro was built. I threw out the decoder, rented GPUs, trained a small adapter to replace it and fed that into a pretrained vocoder.

From under 2x to 15–20x realtime, on the phone. Around 5 times the best public numbers I could find. It's not ElevenLabs, but it's pretty damn close and costs nothing to run, so listening is free. It even sounds better than Speechify's.

Morph is live on the App Store.

While working on Morph I hit a major issue. I either had to paywall it, get funding, or make money quick. I had about 1 month until I ran out of money.

So on the side I tried making slot games....

I'd never made one in my life. Stake Engine lets outside studios publish on Stake, so I started my own provider, CatCrack. Each game took me 5 to 7 days.

I released my first few games and made a decent amount of money. But more importantly I learnt something that seemingly no other studio had caught onto yet.

I DM'd everyone in the biggest Stake Engine streamer's circle, looking for a way to get him to work with me.

I ended up finding my way into a call with a new provider launching with that exact streamer.

After they saw what I was able to do with CatCrack, we joined forces.

In the first month we have taken Blackout Studios to #31 of 1,779 publishers on Stake Engine, doing $9M+ in turnover. See the numbers.