Hacker News Reader

Comment by stroupwaffle | Hacker News Reader

I watched the video promoting Continue.dev and just think to myself, these AI editors don’t actually add much value.

EMacs and Treesitter is much faster way to “find all references to X” in the codebase. Other questions can be answered with the naive grep implementation and marginal brain power.

If all they’re investing in here is to write boilerplate code, well, that takes not much time in the grand scheme of things. Where the value add is actually in the design phase. And, as a result of these tools people are going to just skip those crucial steps.

Am I missing something?

View on HN

> Am I missing something?

Yes, you are basically posting a modern equivalent of the famous "Dropbox reply". "EMacs and Treesitter...naive grep implementation".

There's value in these tools, that's why GitHub, JetBrains, Cursor all built businesses on top of AI coding extensions. Personally I don't use the "Write this full function for me"-features but use it as a way smarter auto-complete. People don't really use them to jump to "find all references to X" as that's perfectly well solved.

I don’t think expecting someone to know how to set up an editor (or not; there are plenty of turnkey distros like LazyVim) or knowing how to use the most basic of *nix tools is on par with the Dropbox reply.

> there’s value in these tools, that’s why… all built businesses on top of AI

Or is it because those companies realized they would rapidly lose market share if they didn’t? You don’t have to add value to become popular, you just have to make people think they’re missing out. Eventually your house of cards might come crashing down, but in the meantime, you’ve successfully enriched yourself. If you have a competing product that is technically equal or superior, it can be maddening to see the popular kid surpassing you without merit.

> You don’t have to add value to become popular

Nothing in the company I work for got adopted as fast as GitHub Copilot. I think it provides value for myself too. So saying that it doesn't provide value is definitely wrong, and I was skeptical at first too.

JetBrains built a business on top of IDEs long before AI was mainstream.

GitHub built their business on top of Git and collaborative coding. Microsoft (who owns GitHub) built part of their business on Visual Studio (also IDEs) long before AI was mainstream.

So it’s a natural extension to their existing business.

But if you’re selling me on a smarter autocomplete, I can already say from experience that AI can bounce good implementation ideas around, but it ALWAYS takes my intervention to get it right.

I’m happy to pay for an AI service, but I’m not going to pay for an AI service, an AI coding extension, an AI diff util, an AI SCM, and so-forth.

I ve tried to use these editors. I tried using copilot and gemini... They all hinder me more than help. And yet I use chatgpt few hours per day copy/pasting my code between my editor and a browser and it makes me massively more productive.

Why? Because all these editors/add ons overwhelm the model with useless context while at the same time lack actual guidance of a proper prompt. They are all follies giving people false hope they can have stuff "happen" without any skill involved.

> People don't really use them to jump to "find all references to X" as that's perfectly well solved.

Clearly not a software engineer.

My imposter syndrome :(

Reply on news.ycombinator.com