Erin.codes

Erin.codes 👩‍💻 freelance web designer & developer
🎧 build, ship, repeat
⚡️coding progressive web apps & learning as I go
đź’Ś [email protected]

05/22/2026

just trying to celebrate the small wins and the universe humbled me so fast

05/20/2026

GOALS!!!

05/12/2026

It’s the same video….So is the server actually storing every quality level separately?

Yes, but the smarter question is how.

Think of it like ordering a pizza. You can get a personal size, a medium or a large. The restaurant doesn’t make one giant pizza & slice off pieces, they prep each size separately so they’re ready to go the second you order.
Same recipe, different versions sitting in the kitchen.

When you upload to YouTube, the platform transcodes your video into a bunch of versions at different resolutions (144p, 240p, 360p, 480p, 720p, 1080p, sometimes 4K). That’s why processing takes forever after upload.

Each version gets chopped into tiny chunks, usually a few seconds each. So a 10 minute 1080p video is actually hundreds of little segments. A manifest file tells your player where every chunk lives.

This is called adaptive bitrate streaming. Your player is constantly checking your connection speed & grabbing the next chunk at whatever quality your bandwidth can handle. If your Wifi drops on the subway? The Next chunk comes in at 240p. If your Wifi recovers? Then it’s back up to 1080p. You’ll barely notice the switch because it happens between chunks, not mid-frame.

The two main protocols that do this are HLS (Apple’s version) & DASH (the open standard), You Tube uses DASH.

The alternative would be streaming one big file & hoping for the best, which is a buffering nightmare on bad connections. Adaptive streaming lets your player make a new decision every few seconds, brilliant right?

So yes, YouTube stores every version. They just store them in tiny pieces & let your player play DJ with the chunks in real time.

Did you know that?

05/12/2026

Why is Facebook Ads Manager so buggy? Why does my bank app log me out every 4 minutes? Why does the airline website forget my booking the second I refresh? It’s 2026 for god’s sake.

Humans wrote that. Humans with degrees & standups & sprint planning.

Meanwhile claude built me a working app in an afternoon that doesn’t crash when I sneeze on it 🤷‍♀️

So is vibe coding actually bad or are we just mad we got gaslit by “real engineers” for years?

For anyone new, vibe coding is when you build software by chatting with an AI instead of writing every line yourself. You describe what you want, the AI writes the code, you test, you iterate. The term was coined by Andrej Karpathy.

The case against it:
You don’t fully understand what you shipped. When something breaks at 2am, you can’t debug code you’ve never really read. AI makes confident mistakes & if you can’t spot them, they go to prod. Security holes, weird edge cases, performance issues that only show up at scale.

The case for it:
Most software is not the space shuttle. Most software is internal tools, marketing sites, small SaaS products, scripts that move data from A to B. Stuff where “it works & ships this week” beats “it’s architecturally pristine but took 6 months.”

And let’s be real, plenty of professional engineers ship code they don’t fully understand. They copy from stack overflow, they pull in npm packages with 400 dependencies, they use frameworks that abstract away half the work. Vibe coding is just the next step in that same trend.

The actual answer:
Vibe coding is great when the cost of being wrong is low & you can verify the output. It’s risky when you’re building things that handle money, health data, or anything where bugs hurt people. The skill isn’t avoiding AI, it’s knowing when to trust it & when to actually read what it wrote.

Now if you’ll excuse me, Ads Manager just crashed again.

Address

San Francisco, CA

Alerts

Be the first to know and let us send you an email when Erin.codes posts news and promotions. Your email address will not be used for any other purpose, and you can unsubscribe at any time.

Share