Tech Videos Are Getting Longer and Saying Less (And We Keep Watching Anyway)

The 45-Minute Tutorial That Could've Been a README

Last week, I needed to understand how React Server Components handle streaming. Found a video: 47 minutes. Scrubbed through it for 10 minutes looking for the actual implementation details. Gave up and read the RFC in 8 minutes instead.

This keeps happening. Tech videos have gotten longer, more produced, and somehow less useful. The average programming tutorial on YouTube is now 23 minutes—up from 12 minutes three years ago. That's not because concepts got more complex. It's because the algorithm rewards watch time, and creators optimized for it even when the content didn't need it.

The weird part? We all know this is happening, but we keep clicking anyway.

When Production Value Replaced Information Density

Tech YouTube used to feel scrappy. Someone would screen record their terminal, talk through a problem, and upload it raw. The production quality was terrible. The information density was incredible.

Now we get 4K footage of someone's perfectly lit home office, smooth transitions between code editor and talking head, and sponsored segments for password managers. The videos look professional. They also take three times longer to say the same thing.

The shift happened gradually. Creators discovered that longer videos performed better in recommendations. YouTube's algorithm explicitly favors watch time over completion rate. So a 30-minute video where viewers watch 10 minutes beats a 10-minute video where they watch all of it.

The incentive structure is broken, and everyone knows it. But knowing doesn't change behavior when your livelihood depends on algorithmic distribution.

The Timestamp Problem Nobody Talks About

Go look at the comments on any popular tech tutorial. Half of them are timestamps. "3:45 - actual code starts" or "Skip to 12:20 for the solution." This should embarrass us.

Viewers are literally crowdsourcing the editing work that creators should've done. When the top comment is always a breakdown of where the useful parts are, that's not engagement—it's a symptom of structural bloat.

I've done this myself. Made timestamps for videos I found useful but padded. Felt helpful at the time. Looking back, it's absurd that we've normalized this as the solution instead of demanding tighter content.

Text Still Wins for Technical Content (But Nobody Wants to Admit It)

Here's the uncomfortable truth: for most technical content, text is still superior. You can skim it. Search it. Copy code without retyping from a paused video. Reference it later without scrubbing through 20 minutes of footage.

But text doesn't pay. A blog post might get shared on Hacker News and generate a few thousand views. The same content as a video can hit hundreds of thousands of views and generate actual revenue. The economics push creators toward video even when it's the wrong medium.

The exception is when video actually adds value—watching someone debug a gnarly issue in real-time, seeing UI interactions that screenshots can't capture, or following along with live coding where the process matters as much as the result. But that's maybe 20% of tech video content. The rest is text that got inflated into video format because that's where the money is.

The Conference Talk Industrial Complex

Conference talks are their own special category of tech video dysfunction. The format is rigid: 30-45 minutes, slides with code snippets that are barely readable, speaker nervously pacing while trying to remember their script.

Most conference talks would work better as blog posts with embedded code samples. But conferences need content, speakers need visibility, and companies need to justify travel budgets. So we maintain the fiction that this is an optimal way to transfer technical knowledge.

The best conference talks I've seen break the format—live coding, interactive debugging, or deep dives that actually use the full time slot. Those are rare. Most are pitch decks stretched to fill time, and everyone in the room is half-following while catching up on Slack.

What Actually Works

Some creators figured this out. ThePrimeagen makes videos that are basically him reacting to tech content in real-time. No script, minimal editing, pure information density. His 15-minute videos cover more ground than most polished 45-minute tutorials.

Fireship built an entire channel on hyper-compressed content. Videos are 3-5 minutes, every second is useful, and the view counts prove that audiences will reward efficiency when it's actually delivered.

The pattern is clear: audiences don't want longer videos. They want better videos. But "better" doesn't optimize for the same metrics as "longer," so most creators never make the shift.

The Real Problem Is Incentive Alignment

This isn't really about video as a medium. Video can be great for technical content when used appropriately. The problem is that platform incentives and content quality have diverged completely.

YouTube wants watch time. Creators need algorithmic distribution. Viewers want useful information. These goals used to align reasonably well. Now they're in direct conflict, and viewers are the ones paying the cost in wasted time.

We could fix this. Platforms could reward completion rate and viewer satisfaction over raw watch time. Creators could optimize for value over length. Viewers could stop clicking on 45-minute tutorials and demand better.

But changing incentive structures is hard, especially when everyone's optimizing for their own local maximum. So we'll probably keep making longer videos that say less, keep watching them anyway, and keep complaining about it in the comments.

At least the timestamps will be good.