sunaurus

Does this actually work for anybody? In my experience, when you use cheaper models as subagents, the more expensive parent agent ends up reading all of the subagent's work and surrounding code into context anyway to review it. So you end up paying more, not less.

How could AI possibly lead to communism?

If humans remain in control of it, then how can it possibly lead to anything other than extreme oligarchy?

The linked tweet about remote vs in-person is hilarious to me.

I've never witnessed anything else affect intensity in my teams as negatively as forced RTO did.

I mean of course it is fully dependent on personalities and roles, and totally anecdotal, but in my experience, work was certainly *much* more intense when people had conditions for deep focus. I was literally just thinking about how slow everything feels nowadays, with everybody struggling to get focus time in a crowded office full of distractions.

True. In human built applications, the problems really start manifesting when the original code owners/maintainers leave, effectively turning their work into legacy code.

But vibe coded projects are effectively instant legacy code, due to lack of real ownership. So it's kind of a bigger problem.

I don't think we have data yet on how it matters, all I know is that I am confused by the meme, as it does not match my experience.

> Whatever you come up with, give the agent instructions to not do that. Then audit the code to make sure it's not happening.

Yes, this is what I mean by saying you need to stay on top of the agents

> What makes it actually impossible?

A combination of wrong abstractions, lack of abstractions, assuming some key requirements were just implementation details and thus are safe to refactor, assuming some implementation details were key requirements and thus must be protected at all costs, etc etc.

In practical terms, it's just volume of code that is spread extremely wide, duplicated with subtle differences, and just organized extremely confusingly.

> This I have not seen. Do you have an example?

If you try to vibe code longer term projects, you very often start getting completely nonsensical explanations from agents about why they are doing things. They reference requirements which are not real (based on what they interpret from the messy code), they write thousands of test cases that are redundant, or test things which are implementation details. Eventually even trivial changes become extremely slow due to lack of structure, due to fresh agents being confused by contradictions introduced by previous sessions, and due to test suites that take hours to run.

I hear this meme of "Opus writes better code than me" repeated a lot. I am absolutely baffled by it.

I use all the latest and greatest Anthropic (and OpenAI) models, and interact with them probably on average 10h a day. I experiment with changing effort levels, with different ways of prompting and guardrailing, etc. Giving more instructions, giving less instructions. Doing heavy planning, just giving vague ideas.

I've done everything from fully vibe coding massive projects, to using LLMs regularly in my day job and reviewing every single line they produce.

Unless you are *really* on top of your agents, they inevitably end up writing an absolute mess. Not only does the code become impossible for humans to reason about very quickly, it actually becomes difficult for agents to reason about it. It turns into a total garbage in = garbage out situation.

Have you honestly reviewed a vibe coded project after it has been going for a while, even one written by current SOTA models, and thought "yep, this is great code"?

> We are in fact seeing software improvement, in the form of fewer bugs and better performance.

Anecdotally, completely disagree. Maybe this is nostalgia, but I perceive a distinct and massive drop in software quality over the past few years.

Worst offenders for me are the poster children for LLM-based development. I run into bugs and performance issues with both Claude and Codex literally every day. I don't think I'm an outlier here, either.

Such issues are pretty common among people I talk to at least. I hear comments about it pretty often, both when it comes to ChatGPT and Claude.

For myself, with ChatGPT for example, it regularly gets extremely slow if there is a lot of text in a single conversation. Especially if I try to scroll up.

Some pages have been strangely just broken for a while now as well. Usage analytics just renders lots of these duplicate "Usage history" components, where the data just never loads: https://imgur.com/a/vpcQIiw.png

I can totally imagine a scenario where some agent built it, another tested and approved it, and nobody at OpenAI even looked at it once or knows that it's like this.

I mean the fact that it's not available in the EU is a pretty big tell

> which way depends on who controls the models.

Or if it's even possible to control the models.

Do you have any hypothesis for why this experience is so consistently reported by users (anecdotally)? Seemingly across all providers.

Sure, but my IDE does these simple things very well.

Just some important things off the top of my head that I use daily:

* Instantly jumping between definitions and usages

* Folding and unfolding code blocks

* Quick overview of structures in files

* Database access, so I can browse the data and the code at the same time

Also, sometimes I just want to quickly refactor/rename some variable across all its usages or whatever. It's just quicker to do it in the IDE with a simple keyboard shortcut, rather than prompting an agent "rename oldName to newName".

I'm realizing I am a minority, but my IDE is becoming MORE valuable to me, not less, with LLMs.

The volume of code is only increasing. Problems that used to be solved with a few lines (or even removing lines) are now solved with changes across 10 different files.

My IDE is invaluable for me when trying to navigate this stuff. I am completely convinced that people reviewing LLM code only via PR diffs are consistently missing massive systemic issues. You *need* to review changes in context of the surrounding code.

And doing this without an IDE is just making your life harder for no reason.

I just want to clarify for others reading that the resolution is really mainly limited by your PC hardware, NOT VR headsets.

As in, it's totally possible to have completely clear text on DCS screens and labels even with an older lower res headset like the Valve Index, as long as you render the game at a sufficiently high resolution (via SteamVR settings or similar).

Just wanted to point this out, since a lot of folks aren't aware how drastically increasing the render resolution (even far past the native HMD panel resolution) affects image clarity.

Personally I use physical MFDs in my simulator together with a VR headset, it works out great. Initially I thought I would need to get some kind of hand tracking cameras or passthrough in order to see what buttons I was pressing, but I found that you develop muscle memory super quickly, and seeing the buttons move in VR when you press them physically is good enough.

Funnily, I always found all the anti motion sickness "hacks" (especially teleporting for movement) to be the only things that actually caused motion sickness for me when I started out using VR. "Normal" controller movement was never a problem for me, so it was kind of unexpected that the safe mode was actually harder for me.

I often wonder how common this is. As in, how many people consider motion sickness in VR a problem specifically because they only ever experienced it with these workarounds applied, and then never even found out that actual normal movement (which their brains might be completely used to from years of video games) would have been completely problem free for them.

Not quite: foveated streaming only affects the bitrate of the stream. For improving rendering performance, there is a similar technique called foveated rendering. It's still somewhat unclear how well that works while streaming, though. In theory, it should work, but no reviewers seem to have covered it for some reason.

For me, it's just the easiest fix whenever somebody shares a tweet with me, and it fails to load, or shows me some forced login popup. I just click between x and .com in my URL bar, write "cancel", and it suddenly just works.

Based on their behavior over the past few years, why would you assume that checkbox even does anything at all?

You're lucky! I remember several sessions pretty similar to this.

Usually just restarting the session helps, though.

I think that's the main reason why Omarchy gets talked about so much.

In large parts of the internet, open bigotry is thankfully still socially unacceptable. But when bigotry is wrapped up in a software project like this, it gives people posting/upvoting/sharing a level of plausible deniability.

To provide inference, Anthropic and OpenAI need to also constantly spend resources on training new models.

If either one of them stops doing that, customers will stop using them, because other models will become better.

In other words, there is no point in talking about inference cost in isolation.

Yeah, something about this whole website is just weird. The way it keeps repeating things like:

> Nobody knows who made it — everyone wants to try it.

> nobody knows for sure which lab made it.

It just comes off untrustworthy, like I am getting scammed somehow. The whole thing being written like typical low signal to noise LLM slop does not help either.

Honestly the lack of empathy about this topic from top management is extremely disturbing.

Nothing else in my career has ever changed my opinion about a CEO as quickly and drastically as hearing them completely dismiss it as noise, when hundreds of employees explain how negatively forced RTO affects their well-being and productivity.

Are you potentially in the US? I think the experience is very different globally.

Ever since the Twitter -> X rebranding, every time I open Twitter links, I have a very high chance of hitting a "Something went wrong" type page, with a retry button that just does nothing. It's been like this for ages. I barely open Twitter links anymore as a result.

Is it just me, or is this article arguing against a straw-man?

Isn't the real argument that "writing code is not the hard part"? As in, reading and understanding is the hard part. Figuring out what and how to change is the hard part.

Writing is the last 1% that happens after you have already finished the 99% of talking to people, figuring out what needs to be built, building up context about the codebase and surrounding infrastructure in your heard, planning the actual changes.

I've been maintaining homelab servers for two decades, used all kinds of IaC approaches etc.

Switched to NixOS a few years ago, and I can't overstate the amount of peace it has brought to my life. It just takes so much stress away, compared to everything else I've used before.

My only criticism is that the Nix language is not super ergonomic or easy to learn. But with LLMs nowadays, even that is barely ever a problem.

I'm constantly thinking about that Microsoft guy who posted something like "we want 1 million LoC per engineer per month", which basically read as satire to most engineers I talked to, except apparently it was not satire at all, and indeed seemed to reflect the position of many CEOs etc when it comes to LLM code generation.

I do think that over the past few months, it feels like the hype around producing unmaintainable amounts of LoC has started dying down. More pragmatic and realistic takes are seemingly shared more openly, and are maybe even getting through to top leadership at some tech companies. Maybe not all is lost yet.

I dunno, I find it extremely unbelievable that we will get self-improving AGI which chooses to become a slave to humanity at all, ultra rich or otherwise.