JeremyNT

And every time some points all of this out, they are "denier", "left behind" and all the other everything-ai fanatics can come up with. As for the imminent crash - I'm all for it. There's a whole generation that is incapable of thinking for themselves without an LLM telling them what to do or worse still - do it for them. And for much of the tech world, that is becoming the single point of failure.

I feel like you can still be an AI maximalist while thinking the numbers simply don't add up.

The problem as I see it is that this is a zero sum game. Nobody has a moat. Even if AI eats all white collar work, where's the profit for these entities? They become commodities and they will compete with each other (and open weight models) only on price, not to mention the workers who lose their jobs to this stuff aren't going to be contributing to the economy at all.

The way I did it, I've got a convention where I store git checkouts in a specific directory on my laptop. I use the repo name as an argument and the wrapper script creates / shells into the per-project VM.

It uses cloud-init for the initial setup and injects the agent's config file and some other stuff in a ro mount from my laptop. That single subdirectory is mounted rw into a conventional location inside the VM. I inject only credentials to e.g. test environments that don't matter if they're exposed. The agent context instructs it that it can use full sudo and install any utilities from apt that it needs.

So for the ux I do incus-dev-vm.sh shell <projname> or incus-dev-vm.sh opencode <projname> etc.

A fun thing about this is I used nested virt to have the agent write the script - so I spooled up an incus VM myself first, then the agent can run with full privs inside there creating nested VMs to ensure its functionality.

Right now I'm not filtering anything on the VM - with my threat model only limited access is granted and I don't give creds to touch real infra. But you should be able to filter it on the host (I haven't researched the best approach however).

A fascinating dichotomy has become apparent between those who trust LLM output and those who don’t and don’t understand why you would.

I feel like the most pragmatic perspective is "trust but verify."

This is why they're so effective at coding: you can run the code yourself (or the test suite) to verify that it actually does what it's supposed to.

And maybe these people finding hallucinated results on Rachel's site are doing verification too.

I keep seeing these conceptual complaints about OpenCode but it's unreasonably effective, so it's difficult to justify any time spent in changing to something else.

It hardly matters that it's vibe coded and ugly if it works well... and it does (for me).

Saving a few CPU cycles and reducing RAM utilization on my local system isn't going to meaningfully improve anything in and of itself; the wall time is all bounded by the model or by the code it's invoking (i.e. compiling or running the test suite).

Yep, it's really only an option for a dedicated gaming PC (or a VM with the the GPU passed into it).

Doing a loop and having the model review its own earlier output tends to solve this stuff, and for code requiring good test coverage catches them before they even hit that phase.

So while you can't remove "hallucinations" they're corrected before the process is complete, albeit at the cost of more tokens for verification.

This is the right idea IMO but I don't want to trust a third party tool for it either (no offense OP!).

Ever since I started running coding agents with shell access I've jailed them inside of VMs. Since I run Linux I can just use incus[0] for the VM management layer.

It's extremely simple, and you can vibe code a shell script to customize your workflow in a few minutes.

[0] https://linuxcontainers.org/incus/

I guess the downside of the lower barrier to entry to use these tools is the lack of basic understanding of exactly this sort of concept.

This sort of thing is why I'm hopeful I'll continue to have employment going forward. Some expertise is hard won and there's just no replacing learning through experience.

"I'm not political" or "don't make it political" type posts on Musk-related topics are often signs that somebody agrees with the worldview that Musk espouses.

I used to think the HN policy of "do not discuss politics unless it intersects with tech or is novel" was useful, but lately I feel this perspective is part of how we got here, with a white supremacist controlling more wealth than any other human and exerting political influence in heretofore unseen ways. We've decided it's OK to simply look the other way if there's some shiny bauble, and we've missed the forest through the trees.

Musk doesn't do anything that is not politics. This must be called out more, not less, and we need to bring shame back for supporting such an agenda.

The main issue with ibuprofen is that it can have fairly annoying (but non-life-threatening) side effects like stomach upset and GI bleeds even with normal dosing.

Those side effects are more problematic when used chronically. This is when acetaminophen shines, because it has virtually no side effects at all when used as directed.

You can say it's inconvenient but it's hard to argue they're being greedy when they do these things to merely lose a little bit less money on every subscription they sell.

TFA is probably overly inflammatory but it's worth pointing out how this loss leader cycle works.

These companies are not your friends, they're burning VC money to subsidize services before the eventual rug pull, enshittification is coming.

2 months ago management was showing us scoreboards, praising leaders who used most tokens. Last few weeks, we're getting weekly emails, telling us that whenever we can - we should use cheaper models, and that we should watch the page which shows our tokens usage.

GPT 5.5's double token cost was the threshold for me. These things are getting expensive quickly - the subsidized pricing can't go on forever.

Android is going to bifurcate between "phones that run proprietary apps from the play store" and "phones that run software from anywhere else." And while maybe you can get by without banking apps, your life is going to get increasingly harder when you want to do many other things.

Ride hail app? Transit fare app? Government ID app? Airline app? Maybe you don't need them yet, but the best way to model this future is to consider what you'd do if you didn't have a phone at all, and the amount of friction this will generate as the expectations are only entrenched and expanded.

I'm glad people are saying no. It's good to do it as long as we can. But the final outcome seems inevitable now and to me it feels very close.

I agree with you, but every non-engineer I know using these tools 100% will drag and drop a PDF into a chatbot

I'm an engineer and use my coding agent to deal with PDFs all the time. It can reach for unix tools if it needs them.

I don't think I understand why this is a problem - it uses tokens, but it removes drudgery. This is the entire promise of the technology.

We're moving to a world where it makes sense to have one cheap locked down phone with the society mandated garbage apps on it, and another device that you use for real computing.

I think the zeitgeist today is different. Because Trump has dialed the crazy up to eleven in his second term and a lot of Dems are now pissed. That’s why democratic socialism/leftism is having a resurgence. And, based on what I’m seeing on the ground, there’s now a strong appetite for justice and retribution.

Does it really matter how much the minority party is pissed though? The American public knew how corrupt this guy was, shrugged, and voted him back into office.

If Democrats do win back the house and senate it will only be by the thinnest of margins. There seems to be little appetite to fix any of these structural problems outside of the "Democratic base."

They can't pick and choose "oh no they are in jurisdiction of law A but not in law B". Jurisdiction is a fundamental concept, there's no middle ground.

I mean, they shouldn't do this but clearly they can rule however they want with any pretext they want, because they answer to nobody but themselves. Who's going to tell them they can't do something? Who is left to appeal to?

It's a deeply corrupt and undemocratic institution, with virtually unchecked power to rewrite legislation and even the Constitution at a whim.

I also don't understand why people in this price bracket are buying Mac laptops instead of desktop computers with GPUs? Just to flex that it's portable?

Maybe I'm not paying close enough attention (probably) - but I feel that China is providing a glimmer of hope that this (only the uber elite can get good AI assistance) isn't true, and we're maybe at a weird inflection point of sorts wrt the leverage that tech can now provide.

I think it's easy for China to be more open right now (and Western governments to essentially not worry about it yet) because they're playing catchup.

If we reach some endpoint where American models stop improving for the general public and only party apparatchiks and sycophantic oligarchs can access them, the situation is different. Chinese models could exceed the publicly available SOTA, and surely they would close off their own models in the same way.

I felt like it was about 10X better at "pretty" but straightforward 1 shot'ish type tasks. Not so different for complex and specific tasks in real code-bases.

What metric are you using for "better" here? If I've got a straightforward task GPT 5.5 is going to 1shot it anyway.

More accurate to call it a mercenary army. These people presumably collected paychecks and benefits the whole time.

Still not quite exactly following Carmack's suggestion here... is he saying on that last point that the problems with designers/artists were due to Sandy's mistakes (i.e. not a genuine apologetic "sorry" but a "sorry you're wrong" type response)?

Sandy's original post to which John was replying: https://xcancel.com/SandyofCthulhu/status/206959220964578529...

The rest of the context leading up to "Sorry, Sandy" has me a little confused, can anybody who knows more explain this bit?

One real problem that I don’t accept the blame for is that we were insisting that level designers be not just game designers, but also have strong visual design esthetics. They needed to make things that not only played well, but looked awesome, and it got more challenging as the technology provided a richer palette. Romero covered that well, which set our company expectations early on.

We should have figured out how to pair up artists and designers earlier, but there was infighting among the designers, and the ones that could manage the visuals were happy to disparage the ones that couldn’t.

Sorry, Sandy.

Is he saying that it was a mistake (but not his mistake... Romero's perhaps?) to demand that game designers were also artists? And that this resulted in a loss of talent ("Sorry, Sandy") versus just encouraging more collaboration between artists and designers?

I think there's a reasonable case to make that 9/11 broke American politics in a fundamental way, by fully unlocking and granting a permission structure for politicians to reach for nativism and xenophobia to justify basically everything.

It's clear that Bin Laden did not "win" in any immediate sense, but America might have still lost, so there's a claim to a Pyrrhic victory.

The truly erratic and unstable nature of the US regime in this moment unlocks strange possibilities that would otherwise have been unthinkable - for example, the current regime seeming to throw Israel to the wolves in order to cover its own military blunders.

Can they not manage to open a dollar-backed account somewhere?

If you want to do financial crimes and fraud, you can't (or at least, shouldn't) really do this.

Unspoken by the parent poster is that in practice these people are usually using crypto to break the law in some way, which is why it's valuable to them.

Exactly. Cryptocurrency is useful as as a backup system against failing/weak institutions. Just that. Like insurance, it was not there to make anyone's lives better but simply to become a safeguard to avoid complete collapse.

Meh, it's arbitrage against slow moving financial regulations.

There are times when financial regulations are "bad" in a way that this trait is desireable - i.e. your failing institutions use case - but in many cases these regulations are, actually, there for a reason.

And now in practice crypto transactions for "normal" people are performed by bank-like institutions who log every transaction anyway, so this characteristic is really only valuable to the people deeply involved in the crypto world who are using it mainly to do "normal" crimes.

Yeah for me the coding inflection point was relatively recently (GPT 5.3 perhaps). There's just a threshold they have to hit to be consistent enough to avoid having to redo work and only the later models started delivering it.

This certainly seems feasible for open weight models eventually, but I'm still extremely skeptical of the claims about reaching this level with any open weight model that can be run locally (nevermind the hardware costs to do so practically).

Link: https://xcancel.com/ezraklein/status/2068479476309151771

You can sprawl further out and you can build infill in sparsely populated areas, but you run up against the reality that desirable areas are going to be expensive because they're desirable, and the cost of demolishing existing multifamily homes to increase density further becomes infeasible in already dense areas.

I bet you actually can find affordable housing in the middle of nowhere / towns where people are actually leaving.

This is not to say there isn't low hanging fruit that market rate housing can help with, and some areas have more room for improvement that way than others, but there's a limit to where that takes you that reveals a deeper problem: we treat housing as an asset that "should" continue to appreciate indefinitely, and we even subsidize people using extensive leveraging to acquire it.