Gabriele Bolognese

The Hacked Channel

It all started when I was 11, in 2020, in peak COVID. The world was locked inside, and I did exactly what every kid my age did: he played games. Minecraft specifically. I tried others, but nothing stuck the same way, I was genuinely addicted, and I didn't fight it. A few months in, I started watching more and more gaming videos, and something clicked: I want to do this. So I started a YouTube channel. I had absolutely no idea what being a content creator actually meant. I genuinely believed that whatever I posted would go viral, no script, no audio quality, nothing. I was posting videos on average daily, some days hitting 4 uploads. On an HP laptop with 1TB of storage and 16GB of RAM. The audio was… bad. Terrible. After a few months of posting raw recordings, I started noticing that people were adding effects to their videos. I was 11, in middle school, zero clue what video editing was. So I did the obvious thing: Googled "free video editing tool." First result was Shotcut. It turned out to be perfect for my machine, lightweight, relatively fast to learn, even if I had to figure most of it out on my own. Around that same time, I hit a wall with Italian tutorials. Searching in English just gave better results. My TikTok feed started shifting to English Minecraft content too, and without even trying, my English improved. Not polished, not British, but real, conversational English. Before I even turned 13, I switched Minecraft's in-game language to English. By the time I was 13, I'd been in the game for over a year and a half, editing for almost as long, but "editing" really just meant cutting clips and slapping on default effects. Still, somehow, the channel crossed 1,000 subscribers with 450+ videos. Looking back now, I'm genuinely stunned it got that far given the quality. The real shift came when I wanted to make YouTube Shorts. Shotcut didn't have a vertical preset, so I had to switch editors, and that's how I found CapCut. This was way back, version 2.9 or 3.0, when auto-captions were still free. For the OGs, you know what that means. Through Shorts tutorials, I started actually learning about pace, hooks, and what real editing looked like.

"The channel was called playmoj"

Then came the crash. The channel had over 600 videos. I was running a hardcore world series, like every other YouTuber at the time, and sitting at around 1,560 subscribers. I was 13, I wanted to grow faster, and like an idiot I ended up on one of those scam websites promising subscribers. They hacked my account, deleted every single video, and replaced them with their own scam content. The channel, playmoj, was gone. I couldn't sleep. I'd lost everything I'd spent over a year building, video by video. I had some clips saved and created a second channel reposting them, but it got 5 views per video. So I quit entirely. That copy channel still exists if you're curious, just don't expect any quality lol: https://www.youtube.com/@playmoj_/videos That hurt more than I expected a YouTube channel to hurt. But it wasn't the end.

First Real Money

After quitting, I didn't touch editing for a while. It was late 2022, I had just turned 14, and I was just browsing around, trying new games. Then I came across Cyberpunk 2077. It looked insane. I had to play it. But my HP laptop couldn't run it, not even close, so I needed a new PC, which meant I needed money. This was the Andrew Tate era. I was a bit late to it because of my age, but after a week of watching dropshipping videos, I stumbled onto one about making money as a video editor. Something clicked. I already knew how to edit, sort of, so I went that route. I created my first Fiverr account and gig. I remember how ridiculously long it took just to set everything up, the gig settings were a lot to go through. But eventually I launched my first public freelancer profile. Worth noting: I was still editing in CapCut at this point. 2 weeks later, a guy with the nickname "hansa" reached out about an opportunity. The job was simple: take a GTA gameplay recording and cut out the best parts. The pay was $5 per video. Each video took well over an hour because the raw footage was sometimes 3 hours long and I just skimmed through it manually. Not exactly great math. Over time I pushed the rate to $15, then $20 per video, which made the hourly pay at least decent.

The channel was Emerals, my first ever client: https://youtube.com/@emeralsrp

Then high school started, and everything got harder. The workload quadrupled overnight and the quality of my edits dropped noticeably. They saw it. But hansa defended me. He knew I was a newbie, he understood the situation, and he kept me in anyway. Some days I was so exhausted from school that I'd edit for 20 minutes with basically zero effort and call it done. He defended me even in those moments. That meant a lot.

The First Crash

In April 2023 I got my first card, a Revolut account under my mom's name, because PayPal wasn't cutting it anymore. By the end of it all, I made around $1,200 from that single editing job over 6 months. It wasn't a lot. But it was mine, and it was the first time I'd ever made real money from something I built myself. June 2023, Emerals quit. Hansa was gone, and just like that I had no income again. But this time was different. Those 6 months had actually made me a decent editor. I knew what I was doing now, at least compared to the CapCut kid who was cutting 3-hour GTA recordings for $5 a pop. And honestly, if I ever become someone, a big part of that is on hansa. He kept me when my quality was trash, when I had zero experience, when school was destroying me. He never had to do any of that. I never even learned his real name. So losing him was my first real setback.

I posted 2 or 3 new gigs on Fiverr. All of them flopped, for months. Nothing. And when nothing is happening, it's very easy to just… stop. So I did. I went back to gaming, no editing, no channel, just wasting time. That stretched all the way to August 2023. Then I came across a channel called "HowToAI," which has since been deleted by YouTube for repeated copyright strikes. That channel was the peak of the YouTube automation era, the idea that you could run a faceless channel using ChatGPT and AI tools and just print money. I didn't know much about ChatGPT at the time, but this channel made it sound like a goldmine. I was convinced enough to actually buy their course.

I started new channels in the video editing niche. And I burned money like wildfire, on YouTube editors, thumbnails, assets, everything. I had zero financial responsibility. The $1,200 I'd worked 6 months to earn started disappearing fast. And then it was just gone. It was December 2023. I had just turned 15. Months of effort, not a single dollar made from the YouTube automation thing. Complete failure. But it gave me something: it showed me that AI was powerful. Like, genuinely powerful. That stuck with me.

Over the Christmas break I started going deeper into it, and somehow that path led me to coding. School hadn't even introduced it yet, that was coming the next year. So on January 3rd, 2024, with a newly released AI called Claude, I started building a tool. January 3rd, 2024. I sat down with Claude and started coding. I had never written a line of code in my life. And Claude at the time was nowhere near as capable as it is now. But I was so excited about this idea of AI just building things for me that I was completely blinded by it. I genuinely thought I had found the cheat code to everything.

I developed for a full month straight. And at some point I needed credits to keep going, which meant I needed money, which I didn't have. So I asked my cousin for a loan. He was 14, running a Vinted reselling operation and selling other stuff online, while I was sitting there with nothing. Which is genuinely funny, because just months earlier it had been the complete opposite. He said yes. I borrowed €200, he sent it over on Revolut, and I bought the credits. But what was I actually building? Let me go back a second.

Remember when I said high school hit me like a wall? The workload was almost unbearable, and on top of that I was doing karate. Time was the one thing I never had enough of. So a thought started forming: what if there was an editor that could make animations faster? Easier? Or even better, automatically? The idea kept growing. I started thinking through every possible use case, every angle, every type of person who might need this. I meditated on it for a while. And then I came up with… The idea had a name: Vision AI Demo. And it had to solve 3 specific problems at once. First, it had to be easy to learn, not overwhelming from the start like so many editors I'd tried before. Second, it had to be completely free, like Shotcut was for me, because I desperately wanted After Effects but couldn't afford it. Third, it had to be lightweight enough to run on any machine, including the kind of HP laptop a broke 15-year-old uses.

So I started building it on bolt.new. Pure vibe coding, zero expertise, just me and an AI figuring it out prompt by prompt. And somehow, it was actually working. Something real was taking shape. Then I asked the AI to move some files around. Reposition them in the folder structure. Simple enough request. It broke everything. Every single path in the codebase got scrambled. Pages were importing files that had been moved or didn't exist anymore, and I had no idea what was happening because I didn't have the technical knowledge to even understand what "file paths" meant. I just knew the whole thing was red and crashing.

Four Rebuilds

I had roughly $55 in credits left. And I did the only thing I knew how to do: I kept asking the AI to fix it. "Fix this error." Over and over, with no further context, because I didn't understand that the root cause was the file restructuring I had requested several prompts earlier. I burned message after message chasing individual errors without ever touching the actual problem. Each fix broke something else. Permanently red. 40 minutes of this and it became clear: it was over.

The project was gone. All those hours. And most importantly, the $200 I had borrowed from my 14-year-old cousin. I had just enough credits left to scrape together a landing page using the few video clips I'd recorded before the crash. It's still up if you want to see what could have been: https://vision-ai-home.netlify.app This one hit like a truck. Much harder than losing the YouTube channel. That was just time. This was debt, something I had never experienced before. It was around March 2024, and my grades collapsed. Zero motivation for anything. Some crying nights, alone, asking ChatGPT why it hurt that much.

The answer it gave me was actually pretty simple: when I started building with AI, I felt something like winning the lottery. I couldn't sleep at night, I'd stay up messaging the AI just to see more progress, more features appearing out of nothing. I genuinely, deeply believed I had won. That I'd be selling this app soon, that the money would come, that everything was about to change. And then it all came crashing down, from the stars to the floor, in a single night. I remember it was a Friday. The moment I realized everything was gone, I had to go to a pizza party with my family. I had to sit there and hide it. What I didn't know yet was that this was only the first of 4 crashes. The first of 4 full rebuilds from scratch. The months after were rough. I tried to restart editing, but there was zero motivation, barely enough to get out of bed some days. I had to force myself because the $200 debt was real and heavy, especially with zero income at 15. $200 at that age isn't nothing.

Nothing came from the editing attempts. Because you can't force output from an empty tank. After about 2 months things started stabilizing, mostly because I landed a one-time gig for $40. Not a lot, but it was necessary. It chipped away at the debt. And from there, I did something risky: I went back to my cousin and asked for another $200 to retry the site. He didn't have it this time. So I had to find another way. August 2024. One year after the YouTube automation disaster. My cousin and I started selling AirPods, not exactly the most ethical operation, but I was still 15 with hopes and dreams, so here we are.

The model was simple: fake AirPods bought at $21 a pair, sold as the real thing. We pooled our money, $40 from me, $10 from him, and bought 2 pairs. A week later we sold 1 for $150. With that we bought 5 more. The plan was to split them, 3 for me, 3 for him, but he'd been doing Vinted reselling for a while already and moved product way faster than I did. So I handed mine over to him too. They sold, though never again at $150, more like $60 to $80 a pair. But it worked well enough that with the leverage of that first successful flip, he lent me another $200. This time I couldn't fail. I genuinely couldn't. There was no more runway. Long story short: I failed again. But not for the same reason as before. This time it wasn't 1 fatal mistake, it was the opposite problem: it got too big, too fast, and nothing was ever finished. Here's what happened.

I started building the editor, which you can still see here: https://create-video-editor-ur5p.bolt.host. Then I got excited and started building a blueprint generator: https://endearing-jelly-e95215.netlify.app. Then another AI video editing tool. Then I went back and tried to remake Vision AI: https://symphonious-maamoul-085d5a.netlify.app. Then yet another AI video editor project: https://ai-video-editor-app-i877.bolt.host. 5 separate tools. All vibe coded. All half finished. All disconnected from each other. None of them good. Another $200 gone. But this time, when the dust settled, it felt different. Not lighter exactly, but different. Because I had actually learned things along the way. New experiments, new patterns, a better understanding of what was and wasn't possible. The biggest realization from all that chaos was this: yes, I could build an editor that runs in the browser. But no, I could not rely on AI to build it for me. That was the line. And I finally understood where it was.

Fast forward a few months. Late 2024, early 2025. Things were stable, not exciting, just stable. School was average, training was consistent, life was moving at a normal pace. Then January 1st came around, and I did what I apparently always do on January 1st: I started building something. This time it was a thumbnail generation tool, because I wanted to give YouTube automation one more shot. I had a few credits left and figured, why not. But something was different about this build. Mostly because bolt.new had switched from their own agent to Claude, which by this point was already becoming genuinely powerful. And working with it, something clicked architecturally that had never clicked before. I had been looking in the wrong direction the entire time. I didn't need to build a video editor in the traditional clips-and-timeline sense. What I actually needed to build was a motion design tool.

That realization changed everything. I still had 4 days of Christmas break left. So I did something I'd never done before: I sat down and started learning TypeScript from scratch, alone, no course, no classroom. I was 16. I started with JavaScript for about a week, then went straight into TypeScript and React. All day sessions, sometimes running into the night. While learning, I also started building. And I figured out that the easiest entry point was a UI design tool. I needed a name, didn't think too hard about it, and landed on FlashFX. Fast, effects, like After Effects. Simple. The very first prototype is still up if you want to see where it all started: https://tubular-profiterole-10ec69.netlify.app. Then go look at what FlashFX is today: https://editor.flashfx.app. The distance between those 2 things gives me actual nostalgia.

Still, I couldn't build everything I envisioned with AI alone. So I also made a separate version myself, something I called FlashFX Editor, a real attempt at building a video editing tool with my own hands. The skills weren't there yet. I knew it. But I was closer than I'd ever been. February 2025. The AI credits were gone. No more bolt.new, no more vibe coding. All I had now was TypeScript and whatever I could figure out on my own. So that's what I did for the next month and a half. Mini projects, one after another. Task managers, small video trimmers, audio remixing tools on the web. Dense, unglamorous, necessary work. I basically learned everything I know in TypeScript during this phase. I'll skim through it because honestly there's not much to say: I woke up, went to school, came home, did TypeScript, went to sleep. That was it. Every day. One casualty worth mentioning: I stopped training for a long time during this stretch, and lost a good chunk of the physical progress I'd built up.

Then April 2025 arrived with a surprise. Out of nowhere, I got a new editing client. Completely random. The month before, in March, I had updated my Fiverr gig presentation video, not because I was actively pushing it, just a small refresh. I had kept the account alive through everything, even through a ban incident I'll get to later. This new client was Italian, working in the financial trading niche, and the offer was actually decent: $250 per month for 4 videos per month. For what the job required, that was good money. The problem was me.

I was still deep in the TypeScript learning phase. I was also trying to actually perform at school for once. And I had just restarted training. The exact problem that made me want to build the software in the first place was now staring me in the face: I had no time to edit. So the edits were trash. After the first payment came in, I was already late on the first deadline. Revisions piled up and I didn't have time for those either. He had paid upfront, and I want to be honest about what I did next: I burned all $250 the same day. Maybe even the same hour. I bought AirPods Max for around $145 to try reselling again, and threw the rest into AI credits. All of it, gone immediately. He threatened to send me to a debt collection agency. I spent an entire day begging him not to. We eventually reached an agreement: instead of cancelling, he wanted 6 videos delivered, not 4, all by a fixed deadline. So I did them.

The revisions were brutal. I think I burned a few hundred hours across those videos. Some had 5 revisions, some had 8. At times 2 rounds in a single day, constantly exporting, sending, waiting, fixing, repeating. My grades collapsed again. So badly that I actually didn't pass the year in June and had to take a recovery test in September just to be admitted to the next year. But the client, to his credit, understood what was happening. He stopped at 5 videos and let it go. When that was finally over, the relief was genuine. I just sat there and breathed. This is probably the center of the entire FlashFX story. After everything, the wasted credits, the 145 fake AirPods Max sitting in a box that I never sold to a single person, the endless revision cycles, the failed year, I was at zero again. And then I saw it: Bolt's world's largest hackathon. Over $1 million in prizes. First place: $100,000 in cash.

That number hit me like a defibrillator. If I won this, or even placed, it would fix literally everything. The debt, the credits, the pressure, all of it. In 1 shot. The excitement that came with that realization was unlike anything I'd felt in months. So I made a decision: full rebuild. Third one. Forget the thumbnail design tool. Forget the motion graphics direction. This hackathon had a lot of design tools competing, so I needed something original. I went back to the editor concept and started building the second version of FlashFX Editor from scratch. And because the hackathon required all projects to be built in bolt.new, I was vibe coding again, but this time with a critical difference: Bolt gave all participants free premium access, around $20 in credits. Every single prompt had to be perfect. I couldn't waste one. I also did something I hadn't done before: I built a team.

There was me. A friend of mine. And a guy I met directly through the hackathon, who went by the Discord username "developer guy." He was in his 30s, an experienced TypeScript developer, while I was still 16. When we didn't want to burn bolt credits, he would jump in and write code directly. He was a massive asset. The build lasted a month. Tailored prompts, careful decisions, developer guy filling the gaps. But when the deadline arrived, the product still wasn't complete. Not even close, honestly. Then the last 3 days happened. Developer guy clutched. He tore through the worst bugs, the ones that made the app genuinely unusable, and got the export function working. Trash quality exports, but exports. It worked enough to submit.

"I finally understood where the line was: I could build it. But I could not rely on AI to build it for me."

June 30th, 2025. Submission day. We sent it in with full hope that something, anything, would come of it. You can still find the hackathon build here: https://flashfx-project-fina-yfbd.bolt.host. Probably the single most important artifact in FlashFX history. And it was also here, somewhere between the last-minute bug fixes and hitting that submit button, that we made it official. We weren't just building a tool anymore. We agreed to make a company. We called it FlashFX.

The decision day for the hackathon winners were on july, i remember attending the awards, full of excitement for the third time, and for the third time, a massive delusion... not only we didn't win anything, but the idea that won the first place was literally a video editor like ours. the idea of an ai video editor has won, but not ours, because it was uncomplete. and once again, out of HUGE, immense regret, i deleted the project, keeping only the link above as final proof. but one thing was learned: we pitched the idea to all the other hackathon members, and it got huge respect. people actually wanted it, there was market demand. this wasn't a sign to stop but to keep going one more time.