> and certainly aren’t obligated to provide it for free
And I'll stop you here. It's less than obvious that there's no obligation. If you provide a critical service that folks rely on at a price less than your cost, you drive out competition, and it's a critical part of your own business model, dropping the service without warning is IMO on the border of what Google should be allowed to do.
It does feel like a lot of very intelligent people here basically start at a first principles belief in property rights, and discover or dispute all of the rights and protections put in place over centuries to patch up the issues that occur when that philosophy meets reality. It reflects poorly on our education systems that these apparently weren't covered or were unconvincing when presented. Or maybe it's just a reflection of the era? In practice organizations seem to be repealing these protections through limited interpretations or loopholes, so maybe that skews people's expectations?
There is a lot of information, in various forms, on the internet that are specifically designed to misinform those who hadn’t taken a course on that particular topic, but leaves the reader feeling they learnt something. Right now LLM’s are good at picking those apart for the reader if they decide to dig deeper, however, I fear this era might not last.
> LLM’s are good at picking those apart for the reader if they decide to dig deeper. I fear this era might not last.
Yeah, I'm not sure that pinning one's hopes for a better-educated populace on LLMs is going to pan out well. Education requires trust and active defense against malign actors.
It's not a poor reflection of our education system, it's all just motivated reasoning. Smart people will move heaven and Earth to argue themselves into a belief that their self-serving position is actually borne of some global altruism.
That our education system wasn't resilient to that well-funded propaganda machine is what reflects poorly on it. That such a machine is allowed to exist reflects poorly on our institutions more broadly. I'll never blame human greed. Systems are designed for humans, if they fail to account for human nature then they're bad systems. I'm not really interested in litigating whether humans as a species are bad.
>Systems are designed for humans, if they fail to account for human nature then they're bad systems.
Systems will always be bad. It's why corporations will always be bad. The complexities are too much for humans. You will never account for all variables. Account for one, with that you are exposed to another. This becomes clear to me when you look at government and the systems it tries to use, since forever. Climate change is another great example. Requires coordinated change across the globe. Many many many factors why that will never change. Change in the system of that size is too hard. So is it the system that is bad, or maybe it's just a reflection of limitations within us as a species, today?
A terribly defeatist attitude. The same could be said about, say, death during childbirth. For hundreds of thousands of years people tried methods of midwifery to ease that process and reduce deaths to little effect. People considered that to be women's lot, an immutable fact of human nature. Then we figured out how to reduce deaths during childbirth to a relatively tiny fraction of all-cause mortality, and that level of care became standard, at least in parts of the world. Why would you be so convinced that systems of organization are unsolvable? Where is your hacker's spirit?
Fun little exercice: How is education funded (not just school, the rest as well) ? What does the salary scale look like ? Would you jump into that boat if had the qualifications ? (and probably: why haven't you jumped into it until now ?)
Once you've got through all of that, how resilient do you expect the system to be ?
Human systems have a critical bottleneck, it's run by humans. That doesn't mean it's necessarily a flaw, but it means all systems are corruptible if it's run by corrupted humans.
And I mean this for any sort of system from corporate, nonprofits, dictatorships, oligarchs, and democracy. Democracy is still a human-run system and that people seem to think democracy is somehow this bastion of freedom is a delusion.
If we want better systems we need better people running them, but that's a conversation that's emerging so we'll see how it goes.
They also happen to be designed by humans, and if you're just begging to have the system fix people's beliefs about corporate greed for you but don't think people themselves are at fault I have no idea why you'd think the systems would be fixed.
Always these complains about corporations or systems or institutions, the responsible person is never "I". If you're unwilling to take responsibility for your institutions why do you think they'd fix your problems? The beauty is people always get the institutions and rulers they deserve, it's not some mysterious system that allows these things to happen, it's you and I.
This doesn't sound like a meaningful critique. You're basically arguing for a culture-first approach to a systemic problem, but insisting that that culture should be one of individual responsibility. I contend that it's exactly that culture that divides the oppressed and justifies exploitation. You've decided a priori that people get what they deserve. I see injustice and try to spread understanding of how our systems create that injustice in hopes that people will change these systems to rectify them.
I'm not at all opposed to the concept of personal responsibility and accountability. In one's personal life it's important to be responsible for yourself. It's also important to understand the context you exist in, and how your actions affect others. It's bad to, say, litter on the streets, and I'll reprimand someone interpersonally for doing so. But if you live in a world where a company comes by and dumps truckloads of trash into your park every week and your government lets them, no amount of personally refraining from littering or scolding your neighbors will get you a clean community. In this case those who need to be held accountable are whoever decided on the dump-trash-in-the-park policy and whoever was supposed to stop them and didn't, and the only solution is a change of policy and creation of accountable enforcement mechanisms.
I'm not just talking about individual responsibility, but collective responsibility emerges from individual responsibility. You start with yourself, then your family, then your community, then your state, then your country, bottom up.
When the company dumps garbage in the town you don't blame the company, you and your neighbors go and put a stop to it. If you're both individually and collectively indifferent then you indeed get what you deserve. That' not an a priori assumption, that's a logical fact. You either take control and self-govern or you're governed. This idea that education or social life works like McDonald's where you yell for the manager if something broken is pathetic.
Vague complaints about 'the system' or crying for some hero CEO, strongman president or influencer or activist of the week to save us poor souls isn't how a free people act. These are problems that can be solved locally from the ground up. You don't need to wait for 'policies' to change, you and your neighbors drag whoever is responsible for that out, or even organize the garbage disposal yourself if need be.
right-wing ideologies are meant to augment concentrated wealth and power, which means there are incentives for the rich and powerful to create right-wing propaganda machines.
left-wing ideologies are meant to create diffuse wealth and power, which means there's no incentive for individuals to create such propaganda machines.
This is why there are enormous amounts of right-wing media, and almost no left-wing media in America.
So all the media that Trump calls "Fake news" is not-left wing?
> left-wing ideologies are meant to create diffuse wealth and power, which means there's no incentive for individuals to create such propaganda machines.
Maybe this was true at some point.
Actually, I think today the left ideologies are used largely as a front, by the people who just want to "augment concentrated wealth and power". I think these are the truly malicious people, because they hide behind the a large mass of gullible population.
They use these shallow "left" idelology to mobilize the masses, and they are shallow exactly because it have to be relatable to the least common denominator. So no nuance, no balanced perspectives, no risk/benefit consideration. Anything that sounds nice on the surface will do (even when it is truly evil after a moments consideration)...
> the existence of an immense right-wing propaganda machine
The biggest trick corporate oligarchs have managed to pull off is convincing people that consolidated markets are "right-wing". Adam Smith is in the public domain, you can read it for free:
A core premise of the book is basically that competitive free markets are good, antitrust is important and government regulations have a tendency to favor cronies and impair competition.
The cronies, of course, don't actually like competitive free markets, so they pervert this as "government regulations including antitrust are always bad" whenever someone wants to do some trust busting. Which in turn sets up their own misconstruction as the straw man to knock down whenever they want to demonize competitive free markets in order to sustain or create regulations propping up their monopolies.
America's right-wing has never wanted competitive free markets, and has never been represented by Adam Smith, the man who said:
"the disposition to admire, and almost to worship, the rich and the powerful, and to despise, or, at least, to neglect persons of poor and mean condition is the great and most universal cause of the corruption of our moral sentiments."
America's right-wing has always been about enriching the connected and the already powerful. Nothing more.
How can we end up blaming the right wing when the propaganda machine is bigger on the other side and even bigger on the government side It's always someone else
I think the idiocy required to agree with the some of ideas of the "american left" vastly exceeds what is required for a complete lack of self reflection.
Ah yes, the left wing propaganda machine. On one side you have Fox and Newsman, on the left you have what? Hasan Piker's Twitch channel? Zeteo maybe? Who are we talking about?
I've self-hosted email systems for businesses for nearly 20 years. I've actually had far easier times delivering to Gmail/Workspace clients than Outlook. Outlook constantly breaks strict DKIM with some of their protection scanning nonsense for emails that seem to get good deliverability almost everywhere else.
I’d say that if Google suddenly stopped providing Gmail for free, destroying the primary means of communication for billions of people, governments would be justified in immediately nationalizing Google with no compensation.
Corporations aren’t magical entities that somehow exist outside of social obligations and can do whatever they want as long as their own terms of service permit it.
> Corporations aren’t magical entities that somehow exist outside of social obligations and can do whatever they want as long as their own terms of service permit it.
Where's your support for this statement in the law?
The existence of law itself is the only necessary support... Law is merely encoded social obligations that the government will enforce. That a single law constrains corporations in any way (and that is clearly the case) proves the statement.
In the broader context GP is clearly advocating for what the law should be, or should be changed to should certain events come to pass. Demanding support in existing law for a proposed change in law is nonsense if that's what you meant to do instead of narrowly discussing the nearly vaccuously true quote you pulled out.
When push comes to shove, the law stops mattering, every time. That’s true for individual rights and it’s true for corporate entities too. The era where things like that don’t happen is a very small slice of human history that is currently coming to an end in real time all around the world. Not long ago, a government simply taking over a company was something that occurred quite regularly.
I have to agree. You can keep pulling that logic back another step (and that seems to have been happening for many steps now) to the point that you no longer have the ability to use the computer.
This can't be dismissed as "slippery slope" logic either. Should elderly people with a bank account be allowed to use a computer? They might read something online and give their savings to a scammer. Frankly, that's a far more convincing argument than the one given here. There's only one solution if your objective function is exclusively to minimize the possibility of a security incident.
> I sometimes get the impression that many people claiming that local models are as good as frontier models work in "token poor" environments. If you can't build large-scale programs using at least Opus 4.5+ then it's difficult to compare.
I sometimes get the impression that people posting comments on HN don't realize that LLMs do more than vibe coding.
Yeah no kidding. For instance, if you are an independent inventor trying to write a patent while keeping your patent lawyer expenses to a minimum, you want to write as much of the first draft(s) of the patent as you can yourself. (You’ll save billable hours with your patent lawyer, and you’ll end up with a better patent because you’ll communicate your innovations more clearly to your lawyer.)
However, and this is the big thing, you absolutely do not want to be asking a SOTA LLM for help with the language in your patent application. This is because describing your invention to a web based LLM could be considered a public “disclosure” of your invention, which, (after a one year grace period goes by), could put your invention in the public domain, basically—and thereby prevent you (or anyone else) from being able to ever patent the invention. Plus, you know, a random unscrupulous employer at the SOTA company could be reviewing logs and notice your great idea, and file a patent on it before you do, and remember, the United States patent office went to “first to file” in 2013.
Oh and don’t take legal advice from random people in the internet by the way.
> This is because describing your invention to a web based LLM could be considered a public “disclosure” of your invention, which, (after a one year grace period goes by), could put your invention in the public domain, basically—and thereby prevent you (or anyone else) from being able to ever patent the invention.
This is simply not true. Even if it were true (and again, it's not) you could simply use zero data retention APIs.
No one at the big model companies is trawling through your chats to steal your patents. It's not only illegal and against their own terms of service, but these people have better uses of their time.
If a competitor to your business discovers that you used a free online AI to help draft your patent 1.5 years ago, that competitor could then cause your patent to be invalidated, which could be greatly to their benefit of course.
The Terms of Service (ToS) for Open/Public AI (e.g., free consumer versions of ChatGPT, Gemini, Claude) often reserve the right to store your prompts and use them to train and refine the model.
Doing an enabling disclosure of your patent draft to another party that is not bound by a non-disclosure agreement is a big mistake, at least while the case law has not yet been settled.
My post was meant to be encouraging to people that might be considering local LLM for this specific use case, where protecting confidential information is of particular importance.
Honestly, I think the SEO virus killed that golden goose long before the first AI chat bot. If we still had good search taking us to sane websites, ChatGPT might well have never been a thing. I was posting (including on HN) about the vulnerability of Google's search business years before AI chat. It just happens to be the thing that filled the gap when usable search disappeared.
Maybe they've decided they don't want to play the same game as OpenAI and Anthropic? They're much better positioned for the high volume AI work that's likely to be where the money is made, with calls to APIs doing routine things for all the businesses of the world. They're also the only big US player that has an open model that you can build on. I don't think vibe coding or the most cutting edge capabilities are what will determine profit from AI.
> They're much better positioned for the high volume AI work that's likely to be where the money is made, with calls to APIs doing routine things for all the businesses of the world
How, exactly, are they currently conquering the enterprise world with their models? What do you think Anthropic is doing?
Their latest proper model is a year old, they have no moat, no enterprise commitment.
Your comment would make sense if they would have actual success in the enterprise market and would have actual products in that area, but they don’t.
They had a brief sprint, caught up, and then dropped the ball again.
Their only current moat is their TPUs, and the fact that
1. The whole (successful) LLM world is screaming for capacity
2. They have excess capacity to rent out, just like Grok
What's a "proper model?" Gemini 3.1 Pro was released 3 months ago. Gemini Robotics 1.6 was released a month ago. And Google is vertically integrated, they aren't just selling tokens, they are selling Taxi rides with Waymo. AI is a lot more than LLMs and Google is doing a lot more than LLMs.
If you're building on top of APIs and can do some eval work (aka do not need the most bleeding edge model), the Gemini Flash and Flash Lite models are super capable for the price.
> How, exactly, are they currently conquering the enterprise world with their models?
I didn't say they were conquering the enterprise world. I said they are better positioned for the work that will be profitable in the future. Winning will mean being "good enough" for things like routine interactions with customers at the lowest cost to the business, and having customers fine tune your models using your hardware.
> What do you think Anthropic is doing?
Aside from being arrogant jerks that don't care about pissing off their customers, they're positioning themselves as the highest price provider for the highest end work. There will be a market for that, and maybe Anthropic will survive, but Google looks to me like they have a shot at being the profitable AI company.
> it is cutting jobs to offset its A.I. spending, saying last month that it would slash 10 percent of its work force.
> Meta also introduced internal dashboards to track employees’ consumption of “tokens,” a unit of A.I. use that is roughly equivalent to four characters of text, four people said. Some said the dashboards were a pressure tactic to encourage competition with colleagues. That led some employees to make so many A.I. agents that others had to introduce agents to find agents, and agents to rate agents, two people said.
Maybe the first to be laid off should be the ones that thought it made sense to track token consumption. Goodhart's Law doesn't even apply in this scenario because that's a dumb metric whether or not you're using it to evaluate employees.
My company did something similar (dashboard to track tokens). It was made available to managers about two weeks before it was available to everyone, so I got to see all my reports' usage before they knew they were being tracked.
The dashboard got announced publicly and just about everyone's usage went up by 100%-200% almost immediately and hasn't come back down, but nothing I'm tracking shows any increase in output since then. We absolutely saw productivity gains a few months ago, but it feels like now people are just burning tokens for the sake of it.
On top of that, as a reaction to the rising costs, we've now gone from unlimited token use to every engineer now having a monthly token budget of $600. I get why that was done, but we're a publicly traded US tech company worth 10s of billions of dollars. We're not hurting for money and the knock on effects are just crazy. For example, I had an engineer in sprint planning say about a large migration type ticket, "Can we hold that ticket until the end of the month? I don't want to burn through all my tokens this early in the month." I just cannot imagine that that's the culture that our executive team was trying to cultivate when they first purchased these tools.
I'm not anti-AI and actually really enjoy using AI for development, but over and over I've watched business leaders shoot themselves in the foot trying to force more AI use on their employees in pursuit of ever increasing productivity. I just keep thinking that there's no way that any productivity gains we've seen from the forced, tracked AI usage are enough to offset the productivity lost from anxiety and churn caused by the unrealistic productivity expectations, vanity metrics, and mass layoffs that have come along with increased AI adoption.
Those executives are simply implementing the directive to inject as much AI as possible into every gear of the economy. Their bonuses depend on this. The idea is that if the world economy becomes dependent on this AI monstrosity, we won't be able to get rid of it. It will be like a situation with a nasty parasite that does a lot of harm, but cannot be removed without the host dying.
That sounds just like Microsoft, Facebook or Oracle products (in some circles). What I'm trying to say it is classic strategy done many times before with various tools (and I hate the perpetrators every time).
Without going too into detail, my company is really, really big on estimates and predictable delivery timelines. An entire years worth of work is speced out, estimated, and scheduled by the end of October the previous year. It's a really terrible process, IMO, but it's the process so it is what it is.
Normally, most teams (mine included) are about 10% behind their plan by the end of Q1. This year my team is closer to 10% ahead despite the fact that we're down one engineer due to a small re-org at the end of last year. These projects were planned and estimated before AI was in heavy use and when, at best, most AI focused devs were still using it like smart auto-complete. Essentially, we estimated the projects before AI was heavily used and we're consistently beating those estimates by a good amount which is not how all previous years have gone.
The AI metrics dashboards didn't roll out until mid-March and, while I'm still seeing us beat our estimates from last year, we're not seeing any additional gains. Basically, all of Q1 we had AI and no dashboards and were beating 2025 estimates by X%. For Q2 we had dashboards and extra pressure to use AI and we're still seeing those X% gains, but no additional gains despite higher token usages.
We also have KPIs around completing a certain number of a certain type of Jira ticket that customers can file and we've seen a similar pattern of an sharp increase in tickets completed in Dec-Feb and then the new rate holds, but no additional increase after the company started pushing AI usage.
It will get really funny when they start imposing an exact number of tokens as a quota, where too little means you are an outdated luddite and too much is inefficient and wastes money
I'm skipping the LLMs/extra steps and simply waiting for the PIP; $EMPLOYER can go ahead and say exactly how many pieces of flair, sorry, tokens they think I need to do my work. It's exactly 0 but I'd like to hear their joke.
Not at Meta but see hints building towards similar madness. War Games, it is. How about chess instead? I'm not big on market manipulation.
A funny Goodhart’s Law parallel showed up in during GPT-5.1 training, where the model was rewarded for using the web search tool, so it learned the behavior of superficially using web search to calculate “1 + 1” and not utilize the result.
> that's a dumb metric whether or not you're using it to evaluate employees
Only if you assume in good faith that the point is to evaluate employees for productivity on some stated goal for the company or role. If you try to view the metric from other possible positions, the one I think fits best is the promotion of token consumption by all means. This is useful for signaling to the broader market that AI is profitable and merits more investment, and may be part of a deal between them and whoever they're buying tokens from. It makes more sense to me that Meta would be more interested in leveraging its control over people to manipulate the state of the world, market, and general sentiment than having them work on stable, well-established and market-dominant software services that really only need to be kept chugging along. Isn't mass-manipulation their whole business? Why wouldn't they use their employees and internal structure to contribute?
I'm reminded of the sales dashboard that tracked the number of calls each sales employee made. There was one employee in 1st place that I assume just always called the same customers multiple times. Her position was about 10x 2nd place.
If someone gave me unfettered access to inference of modern LLMs, there would be no concept of measurement other than the total system wide capacity of whatever the company had available.
It gets worse, my corp is tech adjacent at best, so we get push to use AI, but also get heavily restricted tokens, ridiculous limits on internal tooling ( think context for one short prompt ) and expectation that now one should be able to create the $result fast anyway...
Edit: and if you question that, you are a troublemaker to add to the list
> I think we're in a different world from twenty years ago. The upper-middle class kids I know understand that you can get your hands dirty and that a degree isn't a meal ticket to class security anymore.
This has always been the case throughout my life. I've heard the same thing year after year as long as I can remember. One of the episodes of the Cosby Show had Princeton grads working as plumbers because of the bad job market. What might be different now is comparisons with the job market in the aftermath of the pandemic. New college grads will never see a job market like that again.
> Witness the last two Democratic presidential nominees, Al Gore and John Kerry: one each from Harvard and Yale, both earnest, decent, intelligent men, both utterly incapable of communicating with the larger electorate.
And both running against GW Bush, who attended both Harvard and Yale?
I'm not sure. Microsoft calls Phi-4 a small language model, so the distinction is considered meaningful to some people working in the space. My own view is that the term "LLM" implies something about the capabilities of the model in 2026. Maybe there's not a hard definition of the term, but whatever the definition is, the model in the article wouldn't make it.
"Local inference is rarely cheaper if you’re being honest with yourself about how much you actually use it."
Sorry, but this is not even close to "being honest", it's bad math. That calculation assumes you do nothing with the computer other than local inference.
Huh, you make me curious. Let's actually do that calculation. Let's say you do actually do 24/7/365 AI use. Let's say by some miracle you can do 60 t/s on Qwen 3.6 27b, and let's say this PC cost $3000 (you should be able to do this on a DGX spark, and one of the non-Nvidia models, e.g. the Dell one. $3000 would be a good price, but not totally out of the question). And, of course, let's say these prices remain stable.
So that gets you 1_892_160_000 tokens per year at full blast.
If you go the openrouter, eh, route, you'd get charged $2 per million tokens (anywhere from $2 to $3.6 per million tokens). So the value you'd get from your machine at 100% utilization is 1892 * $2 = $3784 up to 1892 * $3.6 = $6800)
So yeah, not counting electricity and your time the machine "is worth it".
And I'll stop you here. It's less than obvious that there's no obligation. If you provide a critical service that folks rely on at a price less than your cost, you drive out competition, and it's a critical part of your own business model, dropping the service without warning is IMO on the border of what Google should be allowed to do.
reply