> Overall, LLMs aren’t yet at the point where they can replace all engineers. But I don’t doubt they will be soon enough.

All engineers? This doesn't match my hands-on experience at all.

If you give a chainsaw to everyone, it doesn't make everyone a lumberjack. And a chainsaw itself certainly isn't a lumberjack.

If you give Claude Code or the like to everyone, it's doesn't make everyone a highly skilled software engineer. And Claude code itself isn't a highly skilled software engineer.

I've come around to this view. When I first began using these things for building software (the moment ChatGPT dropped), I was immediately skeptical of the view that these things are merely glorified autocomplete. They felt so different than that. These computers would do what I _meant_, not what I _said_. That was a first and very unlike any autocomplete I'd ever seen.

Now, with experience using them to build software and feeling how they are improving, I believe they are nothing more or less than fantastically good auto complete. So good that it was previously unimaginable outside of science fiction.

Why autocomplete and not highly skilled software engineer? They have no taste. And, at best, they only pretend to know the big picture, sometimes.

They do generate lots of code, of course. And you need something / someone that you can trust to review all that code. And THAT thing needs to have good taste and to know the big picture, so that it can reliably make good judgement calls.

LLMs can't. So, using LLMs to write loads of code just shifts the bottleneck from writing code to reviewing code. Moreover, LLMs and their trajectory of improvements do not, to this experienced software engineer, feel like they are kind of solution and kind of improvements needed to get to an automated code review system so reliable that a company would bet its life on it.

Are we going to need fewer software engineers per line of code written? Yes. Are lines of code going to go way up? Yes. Is the need for human code review going to go way up? Yes, until something other than mere LLM improvements arrive.

Even if you assume 100% of code is written by LLMs, all engineers aren't going to be replaced by LLMs.

• nanark

  today at 9:24 AM

  tbh, it could be a breakthrough in model design, an smart optimization (like the recent DeepConf paper), a more brute force (like the recent CodeMonkeys paper), or a completely new paradigm (at which point we won't even call it an LLM anymore). either way, I believe it's hard to claim this will never happen.

  • conartist6

    today at 10:45 AM

    It's pretty easy to understand why it will never happen. AI isn't alive. It's more intellectually akin to a sword or a gun than it is to even the simplest living thing. Nobody has any intention of changing this because there's money to be had selling swords and guns and no money to be had in selling living entities that seek self-presentation and soul

> Our job is to look for AI-native companies (post-LLM startups), because we believe they are a different breed, born in a different world

Is this your firm's investment thesis?

I set the foundation for a Telegram assistant, a web app, and a desktop app, while ditching Figma, Notion, Slack and Pipedrive. Not bad for a fistful of tokens

The amount of bugs and tech debt boggles the mind here

• nanark

  today at 7:30 AM

  hey (author here). obviously, the rest of the content adds a lot of nuance to this statement. it's a bit provocative on purpose.

  But in practice, now that I am working with it, what I needed from those tools already works, with no major bugs so far. I haven’t recreated the tools! just the parts I need to able to plug in and plug out features. Also, many of those features are usually available in great libraries (like Tiptap).

  • anotherpaul

    today at 8:12 AM

    No major bugs maybe for you as a user, but I would bet there are some very serious security issues in multiple components. So I hope none of that code is reachable from the outside. Which ofc is already not true as you are processing data from the outside with llms.

    • nanark

      today at 8:27 AM

      I've been in the game for over 25 years and built plenty of architectures with over millions of MAU, so I am not super anxious about what I've built here. But any API I use can be breached. I would say the risk I take is about the same as using Airtable. And sadly, anything can happen. Also, I am not using it for financial transactions, mostly just deal flow material.

This is a kind of engineering activity I would refer to as "digging your own grave".

My GitHub graph looks the same when I started vibe coding 6 weeks ago.

• nanark

  today at 7:05 AM

  tbh, it's an addiction, can't wait to have a good voice-to-vibecode

  • rmoriz

    today at 7:10 AM

    Armin Ronacher is apparently doing that with some macOS whisper-based app(?)called voiceink: https://youtu.be/tg61cevJthc

    • nanark

      today at 7:17 AM

      wow. on my todo. thanks!

I don't understand why so many VCs fall for "not invented here". Vibe-coded or not, this is just another in-house solution, inferior and more expensive than most out-of-the-box products already out there.

• nanark

  today at 8:35 AM

  (Author here) Maybe you missed the point of what I wrote. I thought the disclaimer made it clear this is just a tiny project for 3 users only and not something meant to scale :) Is my product inferior to Notion, Slack, etc.? OF COURSE. Do I use Notion extensively? Fuck no. I'm more of a Bear (now Craft!) user, but I needed Notion for a handful of tiny features that Tiptap now gives me. So should I pay $60 per seat for the little I need, and miss out on the fun of building my own tool? I think not. But hey, that's just me :)

  • algo_trader

    today at 10:15 AM

    I am more interested in the room (apartment?!!) picture [1]

    Is it a common tourist setup or do people stay there long term? Whats the rent on that?

    Thank god i had company lodging when visiting South Korea

    1. https://substackcdn.com/image/fetch/$s_!kXTm!,f_auto,q_auto:...

    • nanark

      today at 12:10 PM

      it's in Hongdae ("Hongdae T Stay" on Booking). I paid ₩39,000 a night (~23 euros) with a 25 day stay. I don't think people stay long there, it was mostly foreigners. The bed is like a plank, there is no window. But it was cool, I enjoyed it :)

Nice great learnings, Storybook FTW

> Overall, LLMs aren’t yet at the point where they can replace all engineers. But I don’t doubt they will be soon enough, and in the meantime, they just can’t be ignored.

The moment LLM's can replace engineers do we need VCs? Because any one at any time can will any application they want into existence.

• usui

  today at 8:27 AM

  Yes because you can't vibe code social connections or capital.

  • swexbe

    today at 8:46 AM

    Don't really need much capital if you don't need any engineers.

    Don't need social connections if the end-consumer can just conjure up apps themselves with AI.

    • stackbutterflow

      today at 11:01 AM

      If you don't need capital you don't need capitalism.

      How ironic if that's how it ends.

• nanark

  today at 8:36 AM

  I don't think VCs are out of reach. But hey, when It happens, I'll do something else.

> Overall, LLMs aren’t yet at the point where they can replace all engineers. But I don’t doubt they will be soon enough, and in the meantime, they just can’t be ignored

Someone has to maintain that code and there is not a single mention of that caveat after the software is built in the article and 99.9999% of the most widely used software is maintained by humans - even if parts of it was vibe-coded, a human has to maintain it so that it functions correctly, especially a must if it is mission critical software.

It is like we are celebrating mediocrity under the guise of AI and rebranding 'prototyping' as 'vibe-coding' but worse - software with fast accumulating technical debt, slapping on third-party risks and close to no tests at all.

Eventually, someone has to maintain that software and surely 9 times out of 10, a typical senior software engineer will look at that vibe-coded slop and will either throw it all away or reduce these third party services with existing robust open source versions.

Vibe-coding gets you faster to maintain the same negatives from traditional software engineering without you understanding what you are doing.

• nanark

  today at 9:13 AM

  Funny how some people get triggered by this. If you read the whole thing, I actually explain how fucking dumb and clueless LLMs still are once you go past boilerplate. You basically need to keep them in check constantly, even for a pet project like mine. I'm not celebrating anything here, just sharing my journey, the fun and frustrations I've had, going through the stages of vibe-coding: from god-mode euphoria to realizing how deceiving the first rush is.

  But I do believe we'll get to the point they will replace more and more engineers. yes, I don't know how fast, I don't know if LLM will be able to reach that point. But eventually, all that money in research will get somewhere I believe.

  • rvz

    today at 9:46 AM

    > You basically need to keep them in check constantly, even for a pet project like mine.

    Once you do that, then it is not "vibe coding".

    > But I do believe we'll get to the point they will replace more and more engineers.

    Well some software engineers will get replaced. However, your claim was "all engineers", which isn't realistic.

    Given the amount of safety mission critical software that runs the internet, air traffic control for planes, cars and embedded devices, etc they will always need human software engineers to review, test and maintain all of that, including the Linux kernel itself which runs almost everywhere.

    Fully replacing all of them with LLMs would be outright irresponsible.

    > But eventually, all that money in research will get somewhere I believe.

    Of course, LLM security researchers and consultants breaking vibe-coded apps.

    • nanark

      today at 10:02 AM

      > Once you do that, then it is not "vibe coding".

      ok :)