est31

A new generation of AI companies is out there to take over blue collar jobs as well. Check recent YC batches.

Software engineering was a nice target because inputs and outputs are just data and you don't need to figure out robotics. But idk, 3 years ago it seemed illusory (at least for me) that LLMs could take over software engineering, but now here we are. They are still not 100% there yet (software engineers still have jobs), but we are getting ever closer.

Companies are in the process of figuring out robotics, and even if it's not figured out, then we might introduce a gig-ified blue collar economy where an unskilled, underpaid gig worker implements instructions by AI. Plus a lot of blue collar work already today involves robots (cranes, excavators, trucks, etc).

The stable Linux ABI is Win32 provided by Wine.

These deals are part of how the AI economy operates. Amodei has explained this in his recent Patel podcast.

1. Building datacenters takes time. Months, if not years. They take billions of investment.

2. AI revenue is highly unpredictable. Sure, you can make predictions, but maybe your competitor releases a better model 2 weeks after your release, maybe the new model you built isn't as much better, maybe the chinese models steal your show, etc.

3. AI revenue grows a lot. Anthropic's case is 10x per year.

4. So if you are off by just a year in terms of how much GPU you actually need, then that means a 90% of your compute capacity is wasted, and you go bankrupt.

As a solution, companies buy compute from each other! If one company's model did well, they can buy compute from the company whose model didn't do well (like in the case of grok). It's beneficial for both sides, so positive sum game. So deals like this aren't something bad in itself.

It's nothing new either. In SAAS deals, you often commit to a certain revenue and then pay extra if your revenue exceeds that amount. And power market is cut in two as well: longer term deals plus spot markets. Spot prices are way higher than the longer term deal prices.

Given it's SpaceX of course there is financial engineering involved: the GPUs aren't actually owned by SpaceX but a daughter company, and it's been financed via loans that are backed by pension funds. So it's already the case that pension funds back bear the risks associated with SpaceX's operations.

Right now, the bulk of the AI bubble sits in such debt statements and not in public markets.

Yeah, among other factors, that "figure it out" mentality put me off in the end. Especially because often you need to show the same mentality unless you want to overkill proofs and spend more time on them than assigned to you. I sometimes miscalibrated and pointed out some details that didn't need pointing out in my proofs while in other proofs, I skipped over too many details for the TA.

Of course I agree that if the student just asks LLM to do their homework, they have not learned anything. But it's sad if one can't ask questions about a proof or such. Having the LLM around to review the homework submission is also useful, to make sure that the arguments are solid.

One of the reasons why over a decade ago, I dived deeply into the OSS world instead of mathematics was that it was so much more accessible: there were docs for everything, and I got direct feedback when something worked vs when something didn't work. Most of my questions had answers on stack overflow, and once I joined Rust (which back then in 2015 didn't have a big stackoverflow presence) I had a community who answered them for me (and in maths I didn't have that).

AI makes the math world more accessible than before. If you have a question about a proof in the lecture, you can just ask it. Of course, one can't trust it blindly, but fundamentally it's amazing.

I think that's a good thing, but of course this means that a lot has to change in culture and behaviors, also in the research world.

The software engineering world is more or less in the same situation, it's also changing. But for now I think it still holds true that someone who knows maths plus an LLM is better than someone who doesn't know maths plus LLM. At least in software it does.

Horrible to hear this news. Neurological diseases are the worst because we understand so little about them and usually there is no cure, just management.

What have your experiences been with using AI for medical advice? Especially for such rare diseases I suspect that very little shows up in the training data. Personally I'm using AI only for work and only recently started using it for non-work non-coding stuff too.

It's fine to make mistakes, that's how you learn. The problem here was that they didn't announce to the host that they are doing a test of their in-development equipment.

So the host wasn't able to add the additional risk and hassle to the price, which in this instance would have been a quite legitimate ask as the robot damaged their revenue generating property.

It's very ironic that Airbnb itself has done similar practices in the past where it ignored hospitality regulations to establish their business model, i.e. not asking for permission but for forgiveness.

The Airbnb style response would be to gig-ify this model where you ask an independent contractor to buy the test robot, rent the Airbnb, and test it out instead of you doing it yourself. Then the contractor bears the risk of damages to the property.

Well Bezos did actually state that he wants to turn Earth into a natural park.

But yeah, the robot armies don't need grain so why hike up the price of bread? Lack of grain makes those people resentful which means you need to deal with their anger. Sure, it can be dealt with but it's just cheaper to give the humans grain so they are docile. This is basic governance 101 that goes back to the romans (and further).

They also didn't slaughter all horses immediately. You can't eat that much horse meat anyways. It happened piece by piece.

The only good reason for an abrupt mass culling of the 99% (for a coldly calculating rich person with no empathy) would be game theory, i.e. them not being a contender for power any more. If there are no humans, there is nobody who can question the control of the 1%. It would be thus less about economics and more about power.

I am really rooting for the bottom 99%, myself being a part of it, but I really don't know what will happen to us.

If you give it $290 of input tokens for $10 of output tokens, you are doing something wrong. I.e. you paste the whole CI output into the prompt instead of giving it a link to the file, and then the AI greps its way through it (using a fraction of the tokens).

Sometimes AI overdoes things and it re-runs the whole testsuite because the tail command didn't have enough lines, but the other way round messes up the context so much so that in the end all that context is useless.

There is build.rs, proc macros are unsandboxed, and lastly you install the binary so that you can run it. Even if the build and install were fully sandboxed, the binary could still do malicious stuff if ran.

Re vendor lock in point: this is a harness issue really. Sure, CC is restricted to Anthropic models, but it's not the only harness out there. So if one vendor has an outage or botches the quality of their models due to compute shortage, you can switch to another vendor. LLMs are the easiest to switch. Of course, if hardware costs go up, so will all AI vendors. The only way out for the employer would be to directly buy the hardware (or do a fixed price deal with a cloud provider).

Re the understanding code point: you can still use LLMs to understand code. If you write the spec without knowing anything about the code, of course the architecture might suck. Maybe there is already a subsystem that you can modify and extend instead of adding a completely new one for the new feature you are adding, etc.

I use LLMs for my daily workflows and they do understand code perfectly and much more quickly than if I read it.

This is a general fear for me whenever I take a taxi or something like it: i always remind the driver of my luggage in the back when we arrive and ask them whether they can help me get it.

It's unpleasant for me at normal speed settings, but on fast mode it works really well: the AI does changes quickly enough for me to stay focused.

Of course this requires being fortunate enough that you have one of those AI positive employers where you can spend lots of money on clankers.

I don't review every move it makes, I rather have a workflow where I first ask it questions about the code, and it looks around and explores various design choices. then i nudge it towards the design choice I think is best, etc. That asking around about the code also loads up the context in the appropriate manner so that the AI knows how to do the change well.

It's a me in the loop workflow but that prevents a lot of bugs, makes me aware of the design choices, and thanks to fast mode, it is more pleasant and much faster than me manually doing it.

Not a lawyer but the German constitution, Article 12a, speaks of men above 18, not of citizens, or even residents of Germany.

https://www.gesetze-im-internet.de/englisch_gg/englisch_gg.h...

So that article can in theory be used to conscript any man, citizen or not, living in Germany or not.

The Wehrpflichtgesetz, which is a simple law and requires just the 50% Bundestag majority to have it changed, refines this very wide constitutional power in article 1, to require men who hold German citizenship above 18.

https://www.gesetze-im-internet.de/wehrpflg/BJNR006510956.ht...

Article 3 refines it even further to folks below 45 or 60, depending on the severity of the situation.

But yes, in theory it can be changed to include any non-German citizen man, people aged 80, living inside of Germany since a while or never having been to Germany ever, or just random men who happen to change flights at FRA.

This one might last longer. The AI race is on, and the US tries its best to make it as expensive for China as possible to participate in it. Every dollar China spends on GPUs they get at markup is one not spent on building navy ships.

If there is an escalation over Taiwan, then that will cause the loss of most of the world's high grade chip manufacturing capacity. TSMC is busy doing technology transfers into the US, but it is going to take time, those fabs won't have capacity for the whole world, and they still heavily depend on Taiwan based engineers if something goes wrong etc.

Just like with COVID you don't know how long this shortage will last.

It's a loss leader but this is normal. Same has happened with Uber, Airbnb, Amazon, etc. Using VC money to buy marketshare and once you have it, you can milk it.

The question is more around the moats that these companies have and it seems to me while their models are amazing technology, they don't really have a moat. The open/chinese models still continuously catch up to the american ones.

> And how this end is closer with LLMs?

The blog post of this thread argues that now, even average users have the ability to modify GPL'd code thanks to LLMs. The bigger advantage though is that one can use it to break open software monopolies in the first place.

A lot of such monopolies are based on proprietary formats.

If LLM swarms can build a browser (not from scratch) and C compiler (from scratch), they can also build an LLVM backend for a bespoke architecture that only has a proprietary C compiler for it. They can also build adobe software replacements, pdf editors, debug/fix linux driver issues, etc.

If I look around in the FLOSS communities, I see a lot of skepticism towards LLMs. The main concerns are:

1. they were trained on FLOSS repositories without consent of the authors, including GPL and AGPL repos

2. the best models are proprietary

3. folks making low-effort contribution attempts using AI (PRs, security reports, etc).

I agree those are legitimate problems but LLMs are the new reality, they are not going to go away. Much more powerful lobbies than the OSS ones are losing fights against the LLM companies (the big copyright holders in media).

But while companies can use LLMs to build replacements for GPL licensed code (where those LLMs have that GPL code probably in their training set), the reverse thing can also be done: one can break monopolies open using LLMs, and build so much open source software using LLMs.

In the end, the GPL is only a means to an end.

The end game is a resource based economy as all sorts of labor becomes cheap.

Think of Saudi Arabia, Iran, Putin's Russia, or Norway. I.e. risk for highly nepotic dictatorships, with the potential that it might end up well despite the odds (Norway).

Before, if you made a product that improved the lives of everyone, say you invented Google or Heinz ketchup, you could make a lot of money through that, and you did a good deed and became rich the same time. The masses of humans would reward you for delivering the benefits of your invention to them by giving you a piece of their work output.

As their work becomes less and less worth, why focus on those humans though? I am asking rhetorically of course.

An economy that thrives from innovation enriches the innovators, making them powerful. A brute in power causes the innovators to leave or in the worst case, he mass-executes them outright (think of what Stalin did in Russia). With AI, you can have a brute in power though, as an oil rig or datacenter can be protected by a bunch of machine guns.

An economy with AI everywhere will be, after a short and very innovative period, just be about who controls which resource, i.e. water for a datacenter, production lines for robots, mining rights, operational control of robot fleets, etc.

The working 95% will probably experience a sharp decrease in purchasing power, making a lot of products unaffordable to them, so consumption wise we'll have a further shift towards plutonomics. The owning top 10% will probably be affected by this major shift in consumption as well, E.g. a tower full of condos becomes worthless if the tenants can't pay rent because they got laid off, etc.

Need for robots and AI will further increase. Eventually most economic activity will revolve around those robots. It's a bit like paperclip optimizer here, whether those robots protect gay luxury space communism from counterrevolutionaries, or they project the will of the Davos council of Forbes 400, economically it will be quite similar.

There will still be human societies, humans will still talk to other humans. We won't be all exclusively conversing with LLMs, I doubt that. There will still be social mobility but it will revolve around nepotism, lying, and various escalation steps of war.

We might end up in different scenarios depending on the country, but some countries like Germany might lose relevance as most of their value lies in stuff that is going to be replaced by AI, i.e. they have less natural resources, or they have been depleted already.

We might also see companies that automate everything from end to end, from mining to producing and running weaponized robot fleets. Shareholders of those companies will do great too, if the leadership of the companies respects minority shareholder rights that is (why should they though, they will outgun any law enforcement).

Do I like this future? I don't think so. We will probably have solved cancer, communicable diseases, and aging in the next 30 years if AI continues its successful trajectory, but not sure if it will be accessible to 8 billion humans.

You have a lot of control over LLM quality. There is different models available. Even with different effort settings of those models you have different outcomes.

E.g. look at the "SWE-Bench Pro (public)" heading in this page: https://openai.com/index/introducing-gpt-5-4/ , showing reasoning efforts from none to high.

Of course, they don't learn like humans so you can't do the trick of hiring someone less senior but with great potential and then mentor them. Instead it's more of an up front price you have to pay. The top models at the highest settings obviously form a ceiling though.

AI etiquette is a great term. AI is useful in general but some patterns of AI usage are annoying. Especially if the other side spent 10 seconds on something and expects you to treat it seriously.

Currently it's a bit of a wild west, but eventually we'll need to figure out the correct set of rules of how to use AI.

Yeah I meant it in the context of the comment I was replying to, to be precise in the context of the comment that one was replying to, i.e. "10 years of certified Claude Code experience required".

The technology is moving so fast that the tricks you learned a year ago might not be relevant any more.

I think the older AI users are even held back because they might be doing things that are not neccessary any more, like explaining basic things, like please don't bring in random dependencies but prefer the ones that are there already, or the classic think really strongly and make a plan, or try to use a prestigious register of the language in attempts to make it think harder.

Nowadays I just paste a test, build, or linter error message into the chat and the clanker knows immediately what to do, where it originated, and looks into causes. Often times I come back to the chat and see a working explanation together with a fix.

Before I had to actually explain why I want it to change some implementation in some direction, otherwise it would refuse no I won't do that because abc. Nowadays I can just pass the raw instruction "please move this into its own function", etc, and it follows.

So yeah, a lot of these skills become outdated very quickly, the technology is changing so fast, and one needs to constantly revisit if what one had to do a couple of months earlier is still required, whether there's still the limits of the technology precisely there or further out.

I've been 8 years on this site, and I have 8 favorite comments. This comment just made it into a very exclusive club.

yeah it writing stuff that's way better than mine is not the case for me, at least for areas I'm familiar with. In areas I'm not familiar with, it's way better than what I could have produced.

Have you tried the latest models at best settings?

I've been writing software for 20 years. Rust since 10 years. I don't consider myself to be a median coder, but quite above average.

Since the last 2 years or so, I've been trying out changes with AI models every couple months or so, and they have been consistently disappointing. Sure, upon edits and many prompts I could get something useful out of it but often I would have spent the same amount of time or more than I would have spent manually coding.

So yes, while I love technology, I'd been an LLM skeptic for a long time, and for good reason, the models just hadn't been good. While many of my colleagues used AI, I didn't see the appeal of it. It would take more time and I would still have to think just as much, while it be making so many mistakes everywhere and I would have to constantly ask it to correct things.

Now 5 months or so ago, this changed as the models actually figured it out. The February releases of the models sealed things for me.

The models are still making mistakes, but their number and severity is lower, and the output would fit the specific coding patterns in that file or area. It wouldn't import a random library but use the one that was already imported. If I asked it to not do something, it would follow (earlier iterations just ignored me, it was frustrating).

At least for the software development areas I'm touching (writing databases in Rust), LLMs turned into a genuinely useful tool where I now am able to use the fundamental advantages that the technology offers, i.e. write 500 lines of code in 10 minutes, reducing something that would have taken me two to three days before to half a day (as of course I still need to review it and fix mistakes/wrong choices the tool made).

Of course this doesn't mean that I am now 6x faster at all coding tasks, because sometimes I need to figure out the best design or such, but

I am talking about Opus 4.6 and Codex 5.3 here, at high+ effort settings, and not about the tab auto completion or the quick edit features of the IDEs, but the agentic feature where the IDE can actually spend some effort into thinking what I, the user, meant with my less specific prompt.

I still think the source code is the preferred form for modification because it is what you point the AI at when you want it to make a change.

Sure there might be md documents that you created that the AI used to implement the software, but maybe those documents themselves have been AI written from prompts (due to how context works in LLMs, it's better for larger projects to first make an md document about them, even if an LLM is used for it in the first place).

As for proprietary software, the chinese models are not far behind the cutting edge of the US models.

I think it's a complicated issue.

A lot of low quality AI contributions arrive using free tiers of these AI models, the output of which is pretty crap. On the other hand, if you max out the model configs, i.e. get "the best money can buy", then those models are actually quite useful and powerful.

OSS should not miss out on the power LLMs can unleash. Talking about the maxed out versions of the newest models only, i.e. stuff like Claude 4.5+ and Gemini 3, so developments of the last 5 months.

But at the same time, maintainers should not have to review code written by a low quality model (and the high quality models, for now, are all closed, although I heard good things about Minmax 2.5 but I haven't tried it).

Given how hard it is to tell which model made a specific output, without doing an actual review, I think it would make most sense to have a rule restricting AI access to trusted contributors only, i.e. maintainers as a start, and maybe some trusted group of contributors where you know that they use the expensive but useful models, and not the cheap but crap models.

That billion dollar figure is being thrown around for Steinberger's exit to OpenAI, but I couldn't find any reputable source claiming it. It might be a wrong number, idk.

You can't train LLMs on proprietary data, at least not if you want to make that LLM as accessible as Gemini. Otherwise random people can ask it your home address.

So it matters less than one would think. Also, ChatGPT can do 'internet search' as a tool already, so it already has access to say Google maps POI database of SMBs.

And ChatGPT also gets a lot of proprietary data of its own as well. People use it as a Google replacement.