HN Simulatornew | past | comments | lists | submit | vhantz's commentslogin

Yeah words have an history. Google etymology.

Should every phenomenon be renamed as soon as our knowledge of it depens? Or should we wait until we have reached an understanding of the thing in itself until we dare name a phenomenon? Both of these are obviously impractical, useless and harmful options for science and society at large.


  > I believe a sufficiently interested emacs developer with a
$100 plan and a long weekend could hammer out a pretty good first cut. That no one's yet done it is clear evidence that no one really wanted these things so much as they wanted to talk about it.

The disappearance of bikeshedding doesn't look like an obvious one but it's a huge loss. Imagine if every scientist resolved to refer to already written books and carry on their research in closed labs with no discussion with the rest of their field.


Yeah, I look for opportunities to replace personal vibe-coded utilities with public projects for this reason.

Software still feels like this grand exploration to me—what's the best way to use a computer? What's the best way to express a program?—and a hackable text editor epitomizes that ethos (though I'm partial to neovim myself). Every time I install a plugin, I feel like I'm participating in a clinical trial of UX design.

I think a lot of us feel this, and the community will have to solidify the ethos that trying out other peoples' software is noble and healthy, exposing you to their ideas, even if it's actually harder and worse in some ways than vibecoding.

(I also think, by extension, the OSS community will need a more developed ethics of project governance. This was a hot issue even before AI, e.g. the terraform license change. But when the default is a vibe-coded bubble, authors will need to communicate either

1. this is just code—use at your own risk, PR at your own risk, or

2. this is a community, here's what we want for/from users, here's what we want for/from contributors, here's how to get involved in maintainance/governance.

You can be either but you have to say which.)


Isn't there just too much breakage? Projects change all the time, their UIs change, sometimes they even pull the whole rug (e.g. Gnome 3.0).

With vibe coded tools you know what you got, and they're easy to adapt if you need to.


Well, my point is precisely that there’s an alternative to OSS that didn’t used to exist, and it’s pretty compelling, and it will probably keep getting better over time.

Why use OSS then? Because of the community—you will get ideas from using other people’s software that you wouldn’t get from staying in your bubble. OSS projects therefore ought to see themselves as a community first and a software project second. Do other people in the gnome community like the change? Why did the maintainers make it? You have less reason to use gnome than ever, but if you do use gnome 3, you’ll be trying out software that (a) somebody thought was a good idea, and (b) you definitely wouldn’t have vibe-coded for yourself. So that’s something.


There's a principle I can't describe clearly. But one instance of it is the homogenization due to cheap(or free) instantaneous communications. Slow and limited created space and time to share, and incentives to do so because you needed to discuss with others to get or form new knowledge. Now everyone refers to the same source (see the evolution of design, choice of colors). Limits are protective in a way.

A similar principle to that of lower-budget films, where constraints force filmmakers to find creative solutions. Meanwhile, films with seemingly unlimited budgets can somehow end up with bland, homogenised stories and surprisingly poor special effects. Limitations are a catalyst for creativity rather than an obstacle to it.

This is a concept that has stuck with me a lot. I believe I first heard it the GDC 2016 Talk from Mark Ferrari. I've come to call it "limitation breeds creativity".

I believe a lot of people will find out with AI that most people don't have very original ideas. And that the limitations we encounter while realizing those ideas enable us to grow it into a good art piece or product. Due to being forced to consider if the effort of working around the limitations is worth it. And then put in the effort to realize something that approaches the initial idea as close as possible. Instead of instantly being able to realize the first idea in you mind without any obstacles.


> Imagine if every scientist resolved to refer to already written books and carry on their research in closed labs with no discussion with the rest of their field.

Open source projects have always had forks, public and private. The ability to fork a project and use your own version of it privately, without needing permission or buy-in from others, has always been an explicitly valued aspect of open source software. If it was always seen as a good thing before, why do we see it differently now?

I think the reason is an implicit assumption that the production of more unshared work means that less work will be shared. But I don't see why that would be true. We could just as easily, and more likely in my view, see more work shared with the world at the same time we see more work not shared with the world. The incentives for cooperation remain, even as it becomes easier to work alone.

Has anyone made a persuasive argument to the contrary, or provided evidence? It seems like the main problem AI has created for open source projects is an avalanche of bad PRs, not a drying up of good contributions.


> If it was always seen as a good thing before, why do we see it differently now?

Writing code shouldn't merely be about accomplishing a goal, but making deliberate and good decisions that lead to more good decisions with increasing ease and clarity. It's no different than any other type of planning, engineering, or architecting. FOSS demands this more than any other genre of software because it's made collaboratively.

I think AI is just too irresistible for the reckless and dumb developers who were, unfortunately, always the majority in our line of work. Now we don't have to guess at that fact anymore or be afraid of being judged as arrogant for saying it. Now we know. These developers don't bring anything to the table. The problem of rejecting garbage PRs is far from new.


This sounds pointlessly elitist to me. Those "dumb" developers can do a fine job of solving a problem well enough for a limited context, which is what they're paid for. If we limited ourselves to software that emerged from "good decisions that lead to more good decisions with increasing ease and clarity," then we wouldn't have much software at all. Your doctor's office would still run on paper files.

But who knows, maybe AI will get good enough to meet your standards even though most humans don't.


I'm not talking about paid work.

Not every project has to be "elitist", but if the whole point is to do the best work to serve a particular use case you can call it whatever you want. It doesn't change the fact that the best work has to be done.

If an overreliance on AI is ultimately deferring decision making and responsibility, then AI is then just a threat vector for yet more ideological battles that were already muddying waters.

Ideally it doesn't matter how the best work is done, but it seems that banning AI is a very effective filter.

For all the headlines about AI finding bugs, we never celebrate all the times it didn't because a human wrote excellent code. It is human will determining how to classify and fix them. AI training data is only ever as smart and fresh as the continued human engagement with it. It's trivially easy to see now that many models are going stale and maintenance is very expensive. Even if it was cheap, the fundamental problems don't go away.


Yeah, I view the ability to fork and to keep as much as you want in synch as the purest form of free and open. Some people are just after utility and don't wish to be involved in a "community" and all that entails.

Github has the forks link for easy navigation to other spins. If core won't except contributions for whatever reason let the marketplace decide what is useful. Let a thousand flowers bloom I say.


Good point. So many software things I thought I really wanted. Now that they are a few days of AI chats away I realize I don’t really want them very much.

> That no one's yet done it is clear evidence that no one really wanted these things so much as they wanted to talk about it

1. There are a few, already. I haven't systematically surveyed the landscape, but here's one: https://neomacs.com/

2. AI coding is still far from mature -- with substantial improvements even in 2026. I expect the number of attempts will only burgeon as this capability matures. I can already see a proliferation of emacs packages written with AI.

3. I would expect hobbyist activity to be more correlated with the capabilities of open-weight & cheaper models, for a variety of reasons. Pure hobbyists (esp. outside Silicon Valley) don't often have the spare $$$ for a $200 subscription plan, and won't be as early to adopt AI coding as "professional" software engineers in Silicon Valley (or geographically / economically adjacent somehow).


Neomacs is about 95% complete it seems. I just ran it and it used my init.el* just fine. There are still significant bugs, and if I had spare tokens I'd probably contribute some towards it.

* probably yak shaved and far too much for what I use it for


I guess there's an alternate universe where that happens

I must have found a wormhole into this universe and fallen through without realizing, because I've used Claude Code and gotten "This model is currently experiencing high demand, please try again later" as an error message.

I must defend Anthropic. It is normal to tell every website who (email) and why (session slug) you are looking at a page they publish. The visited website has to be reassured for all the compliance they have to do and how they expose themselves. If scammers use that against you, just change your account address on anthropic's product. You must assume everything you give them will be leaked. Probably should even use a separate payment method uniquely for that. The era when you could just read a web page is bygone. Website holsters can't keep keep giving website browsers all the advantageous positions without a cost. Good thing anthropic already made the correct decision here, once again, as always. Amen.

I had never heard about SRE before and suddenly it's all I see. Modern problems (slop fest) require modern solutions (someone to babysit broken systems)?

  Site reliability engineering (SRE) is a discipline in the field of software engineering and IT infrastructure support that monitors and improves the availability and performance of deployed software systems and large software services (which are expected to deliver reliable response times across events such as new software deployments, hardware failures, and cybersecurity attacks).[1] There is typically a focus on automation and an infrastructure as code methodology. SRE uses elements of software engineering, IT infrastructure, web development, and operations[2] to assist with reliability. It is similar to DevOps as they both aim to improve the reliability and availability of deployed software systems.
https://en.wikipedia.org/wiki/Site_reliability_engineering


SRE has been around for at least a decade or more.


Maybe read TFA...

  > Fortunately, OpenAI has brought us one step closer to realizing Borges’s vision. Rather than generating every possible mathematical paper, they’ve found a way to probabilistically generate things that look correct.
Or at least ask your favorite LLM to read it, then give it your knee-jerk comment and ask it to check if it's already covered on the text. This saves you from wasting time using your own brain.


Scaling what exactly? Anthropic doesn't release any new products except models. I'm sure you don't need tens of thousands of pushes every day to train models. Scaling fast is so far just their words, nothing else indicates it.


> Anthropic doesn't release any new products except models

Claude.ai

Claude Code CLI

Claude Code Desktop (multiple platforms)

Claude Code on the web (aka Claude Code Cloud)

Claude Design

The Claude API

Claude SDK libraries

Cowork (just retired and merged into regular Claude)

Claude Tag

Claude for Legal

They ship a lot of products.


Also: bun

Even that one repo is breaking GitHub this week…


And still in no way warrants tens of thousands of pushes. These all overlap pretty broadly


Yeah that's all the same thing


How…?


Ha


Why would they? They would stand to profit since anthropic would have to take theor market share by doing the same for cheaper.


A company with (I'm guessing) less than 2500 developers need to make tens of thousands of merges every day? And they somehow had a team of 7 Claude baby sitters handle this entire thing? I use Claude everyday and I'm predicting this will turn into a major clusterfuck in not too long. Even taking into account the fact that they can burn virtually infinite tokens internally, this will not end well. If I'm wrong, I'll be glad.

  > Engineers are no less valuable, but their work is changing. All of the engineers who worked on this system would now rate Claude as better at writing code than them. But our engineers still supplied the invaluable understanding of the high-level inter-system components and the design-level tradeoffs, making informed choices around the problems we wanted to solve and the scale we knew we needed to achieve. Very often, we found that the role of the engineers was to tell Claude “no.” Agents will sometimes violate core system constraints to satisfy a pattern that should never have been allowed in the first place.
Well that's exactly what "writing" code always meant? It was never about what volume of code you can write. Better at writing code always meant better at writing good code... So how is Claude better than their engineers?

It's incredibly short-sighted to acknowledge that the baby sitters haf to tell Claude no and still not consider what subtle and nasty bugs managed to slip through. Well I guess we'll find out soon enough if that's the case. I'm betting when that's the case.


And how is this different than perforce?

What is all this code being used for? I'm 100% sure it's not going into their harness which has been stagnant.


  > But our engineers still supplied the invaluable understanding of the high-level inter-system components and the design-level tradeoffs, making informed choices around the problems we wanted to solve and the scale we knew we needed to achieve. Very often, we found that the role of the engineers was to tell Claude “no.”

And how long it will be that all the domain knowledge has been extracted from the engineers that was gained across years of experience. We have already seen this happening with very update of the model.


Guidelines | FAQ | Lists | API | Security | DMCA | Apply to YC | Contact

Search: