A developer's workflow for extracting knowledge from tech talks, tutorials, and conference recordings faster than ever.
I need to learn a new framework roughly every four months. That's not a flex. That's the reality of working at a startup where the stack evolves faster than the documentation.
Last month it was tRPC. Before that, Drizzle ORM. Before that, a deep dive into Cloudflare Workers because we decided to move some edge logic. Each time, the cycle is the same. Read the docs (sparse), check GitHub issues (depressing), and watch YouTube tutorials (time-consuming).
The docs give you the "what." YouTube gives you the "why" and the "how it actually works in practice." The problem is that a typical conference talk is 40 minutes, and the 3 minutes you actually need are buried somewhere in the middle.
I found a way to fix that.
The Tool
ChatYT. It lets you paste a YouTube URL and have a text conversation about the video's content. Ask questions, get answers with timestamps, generate notes, export to PDF.
I know how that sounds. Another AI tool, another promise. I was skeptical too. Then I tried it on a 55-minute React Server Components talk from React Conf and got the exact implementation details I needed in under 4 minutes.
Here's my actual workflow.
Step 1: Find the Right Videos
When I'm learning something new, I don't start with "beginner tutorial" videos. I look for conference talks by the library authors, livestreams where someone builds something real, and deep-dive explanations by developers who've used it in production.
For tRPC, that meant Theo's videos, the official tRPC talks, and a couple of "building a full-stack app with tRPC" livestreams that were 2+ hours each.
Total watch time if I consumed everything: roughly 8 hours.
Step 2: Ask Specific Questions
Instead of watching, I paste each video into ChatYT and ask targeted questions.
"What's the recommended project structure for tRPC with Next.js App Router?"
"How does the speaker handle error types between client and server?"
"Does this tutorial use middleware? If so, how is auth handled?"
"What are the performance gotchas mentioned in this talk?"
Each answer comes back in seconds with references to specific parts of the video. If the answer is interesting enough, I click the timestamp and watch that specific 2-minute segment. Otherwise, I move on.
Total time to extract the same knowledge: about 45 minutes.
Step 3: Generate Notes and Flashcards
This part surprised me. ChatYT generates structured notes from any video. Not a transcript dump. Actual organized notes with headers and key points.
For technical content, this is gold. I generate notes for each video I process, export them as PDFs, and dump them into a folder. When I'm coding and need to remember "wait, how did that speaker say to handle optimistic updates in tRPC?" I search my notes instead of scrubbing through a 2-hour video.
The flashcard feature is something I didn't expect to use as a developer. But when you're learning a new API surface - method names, patterns, configuration options - spaced repetition actually works. I review cards during build times. Yes, I time my learning to my CI pipeline. Don't judge me.
Step 4: Multi-Video Comparison
This is where it gets interesting for learning. ChatYT has a multi-video chat feature where you can load multiple videos and ask questions across all of them.
When I was learning Drizzle ORM, there were three different tutorials that each took a slightly different approach to schema definition. Instead of watching all three and trying to remember the differences, I loaded them into a multi-video chat and asked: "Compare how each tutorial defines relations between tables."
Got a clear comparison in 30 seconds. One used the older syntax, one used the new relational query builder, and one had a custom abstraction. Saved me from following a tutorial that used a deprecated pattern.
Real Talk: When This Works and When It Doesn't
This approach works best for conceptual content. Conference talks explaining architecture decisions, tutorials walking through patterns, deep-dives into how something works under the hood. That's 80% of tech YouTube.
Where it's less useful: live coding sessions where you need to follow along keystroke by keystroke, or visual content where the speaker is drawing diagrams on a whiteboard. For those, you still need to watch. But even then, I use ChatYT to generate notes afterward so I have a searchable reference.
Also, the video needs to have captions or a transcript. Most English-language tech content does. Occasionally you'll hit a video without one - just move on to another video covering the same topic.
The Trending Page for Dev Content
ChatYT has a trending section that surfaces popular videos with summaries. I've started checking it for tech content the way I used to check Hacker News.
Found a talk about WebAssembly edge computing there that had 500K views and never showed up in my YouTube feed. The summary told me exactly whether it was worth my time before I committed to watching. It was. But at least I made that decision in 30 seconds instead of 10 minutes into the video.
They also have a blog with some posts about AI-assisted learning that are worth reading if you're into the meta side of how these tools work.
The Numbers
Free to try. 15 credits on signup, no card. Enough to test it on 10+ videos.
I pay $5/month for the Basic plan. Five dollars. I spend more than that on coffee during a single debugging session. Pro is $12 if you need more capacity. A few developers I know who consume a lot of conference content went with Pro.
My Dev Learning Stack in 2026
- Official docs for API reference
- GitHub issues and discussions for edge cases
- ChatYT for extracting knowledge from conference talks and tutorials
- Actual coding for muscle memory
The third one is new. It replaced "watching YouTube at 2x speed while pretending to absorb information." Honest question - how much of a 40-minute conference talk do you actually retain when you watch it at double speed? For me, the answer was embarrassingly low.
Now I retain almost everything because I'm actively asking questions instead of passively watching. It's the difference between reading a textbook and having a conversation with a tutor who read the textbook.
Try it on your next "Watch Later" video. The one about that framework you keep meaning to learn. You know the one. It's been in your queue for two weeks.
ChatYT: https://chatyt.io
Tags: webdev, productivity, learning, programming, ai
Top comments (0)