The Mistake That Cost Me a Year
started building with a regular MEAN stack. i was 17, knew how to code but wasnt experienced, just pushed by those traffic numbers. used yt-dlp as the core scraper along with other fallbacks like youtube-dl and unofficial tiktok apis. by all means i was overengineering, tryign to build a bulletproof system with fallback on fallback. took two months, and by the end changing one line meant touching 15 to 20 files. that's bad code.
rewrote it, stripped external deps, used yt-dlp efficiently in node js. still slow, 10 to 15 sec just to fetch info, another 10 to 15 to start download. tried optimizing args and prewarmed pools, still stuck around 9 to 10 sec. obsessed with speed at this point, 3 months gone.
tried bun, no real improvement, cmd process time was the smae. switched to golang, learned it in 15 days, rebuilt again, and it worked, 4 to 5 sec startup. procrastinated a full week just enjoying that small win.
month 5, optimized further. realized we didnt need full yt-dlp --dump-json metadata (5 sec), switched to oembed and iframely for instant basic metadata (250ms) while yt-dlp handled the download in the bakcground.
launched after 6 months under the name "videosaver," too common a name, google barely indexed it, didnt even rank for its own name. learned naming and domain age matter more than i thought.
depression hit, overthought for a month. month 8, rebuilt frontend and rebranded to "buckty," ai chatbot style ui, technical seo scored 100/100 everywhere. launched month 9, fast and functional, but still zero real traffic. blog posts never got indexed. wrote on medium, other platforms, still nothing. cookies issues and ip throttling kept coming up. tried aria2 for speed, bad call server side. backlinks didnt move the needle either.
started an instagram theme page posting movie clips, hit 50k followers and 70m+ reach in a month. tried converting it into traffic for buckty, started posting memes, and within 3 days meta blocked the account for bot-like activity. never recovered.
realized fast downloads and decent ui isnt a usp. built a mobile app in kotlin with direct-download-with-share like "seal," but downloading server side then streaming to the user. added google drive auto-upload as another usp, connected web and mobile through a nest js middleware sitting on top of the golang backend, handling history, drive uploads, everything. still had the cookies issue. still no audience. decided no more features, just focus on traffic.
spent the last month trying every seo strategy, nothing worked. realized the real issue was leverage, competitors were doing blackhat bulk backlinking, spending 10k a month to pull 200k a month in cpm ads. the other real path is sponsoring a creator with real audience to showcase the product.
after all this, my final takeaway is... nothing, and everything. learned a ton about optimization, design, scaling. but the real lesson: product is nothing, marketing is everything. people building "ai wrappers" get traction i couldnt, even though i built an actual system, not hype.
so i'm open sourcing the whole ecosystem, mobile app, web app, all of it. built to run locally too, using wails to package the web app as a self hosted system app, connect your own accounts, back everything up. designed it to need minimal maintenance, should run fine for 6 months untouched. within a week of open sourcing it, i'm leaving it and moving to the next project. hard to walk away from something like this, but the best time to leave anything that hurts you was the moment you regretted it, and the second best time is now.