With AI-assisted coding on the rise, I'm starting to wonder if humans will still need code search tools, or if generative models will do it all for us. On the other hand, maybe code search will become even more critical, especially as AI-generated codebases balloon and eventually need thorough refactoring, security audits, or just plain manual debugging.
Context: A friend wrote a semantic code search tool (syng.dev) designed for developers maintaining large or legacy codebases, or for web security folks. It's great, but I'm unsure about how demand might shift if coding becomes more "AI-first." That said, I could see these tools being just as useful (if not more so) for the AI agents themselves when sifting through massive amounts of code.
What do you think? Does code search still have a bright future, whether for humans or AI, or is it fated to become obsolete in the era of "vibe coding"?
I once automated the uploading of >26k Gource videos to YouTube. I sorted GitHub by star count, applied some filters (e.g., had to have X commits, Y files), ffmpeg, YT API. That all happened on my little 2015 MBP. Good memories. I'll clean up the repo and share my hacky code sometime. I miss Ruby, what a fun language with which to play.
IIRC, there's another YT channel that took up the torch. I'm grateful they did. I know, they might seem a little spammy, and it can be frustrating when you Google for a repo and only find a mostly useless YT video of a Gource visualization. And yet, random devs used to email me with grateful messages every week. I felt like, in some microscopic way, I was doing good.
Yeah, I need to ponder and write on that. Hm. Well, this bubbly isn't going to drink itself.
I'll edit this comment into shape later. I'm on my phone at Hedwig's.
Love you to the random internet stranger who read this comment.
Love you to the person who posted this Gource link.
Love you to Andrew Caudwell, the creator of Gource. He seems like a good fellow, that Andrew Caudwell. Yes, quite so.
Quickly, before I :homer-simpson-fade-into-bushes:
I think there is much value in Gource visualizations. My remake was rooted in a vocal minority who expressed a few negative comments on my uploads saying they were useless. I suspect they were just frustrated that they were looking for docs/tutorials and only found my video, heh.
With the right options and extensions/augmentation, Gource is incredibly useful. Especially when slowed down and when you add custom tailored config for each repo. That takes time, but can be automated as well. Ooo, that sounds so much fun!! Issue titles, big refactor custom highlights / slow-downs. Have it right next to JSCity (or equivalent for non-JS), gitstats, and... What was that other tool... You know, the one that makes watching network traffic fun? Oh! Codeswarm? No... It was... I'll find it. It's like ping pong. Where is my AI side panel when I actually need info without leaving HN?? Ok bye.
Next time I spam the planet with Gource videos, I vow to do it more justice :cat-salute:
That's some awesome work! I always used to stumble upon videos of yours/similar, partly as it was kind of a way to discover 'niche' but active repos, though mainly because Gource visualisations are just plain fun to watch.
Context: A friend wrote a semantic code search tool (syng.dev) designed for developers maintaining large or legacy codebases, or for web security folks. It's great, but I'm unsure about how demand might shift if coding becomes more "AI-first." That said, I could see these tools being just as useful (if not more so) for the AI agents themselves when sifting through massive amounts of code.
What do you think? Does code search still have a bright future, whether for humans or AI, or is it fated to become obsolete in the era of "vibe coding"?