Posted by dgellow 1 day ago
You're an example of this process. You used to require a frontier model running on a data centre. Algorithms improved and, perhaps, your hardware as well, but your requirements didn't go up. Now you can do what you need to do with local hardware. You're one AI user who might not need to send money to Anthropic going forward unless you find a new frontier application that interests you.
This is why I'm a little bearish when it comes to the valuation of the AI giants. Yes, there are some extremely lucrative things on the frontier right now but, in the future, people may shift a lot of what they're doing now to local hardware while new things on the frontier may not produce the kind of profits that will pay back trillions in debt. AI companies have bet big on themselves being able to constantly and perpetually find new applications, and also that those new applications will make money.
I am definitely building more and more ambitious projects as models improve. Both personally and at work
Did you somehow become incapable of doing things without models? Most of us have not.
The above comment said:
> need to buy tokens from a handfull of tech companies to run their businesses or make their products
That's clearly indicating necessity.
That would indeed be concerning if it was the case, but it's not.
So now the case is just "you can pay more money to do things faster".
And guess what? That's been the case for centuries if not millennia. Nothing special or concerning at all.
And so your analogy is misleading. You're trying to imply that only a tiny number (2-3?) of companies make the tools, and nobody else does. You either pay those companies, or you don't get them.
That's just false. There are hundreds of companies that will sell you tokens - just not for frontier models. The correct analogy is how Samsung, Google, and Apple make the best phones in the world, so you can buy from one of them - or you can buy a still-good but not bleeding-edge phone from another vendor. And everyone is fine with that.
To the leadership of these tech companies that actually is the dream. Sam Altman said the quiet part aloud recently: "From a global allocation perspective, actually, it's quite terrible to have everybody have a bunch of the world's limited supply of DRAM sitting in their own computer, and being used maybe 5% of the time."
And it certainly fits into the neofeudal worldview of other relevant Big Tech leaders. Be it Peter "I no longer believe that freedom and democracy are compatible" Thiel, or unnamed bigwigs discussing if shock collars are effective to prevent their staff from revolting after a doomsday scenario [2].
We shall own nothing and rent everything, that is the mindset these people believe.
(And if you ask me: the AI IPOs are part of that strategy. Assuming OpenAI and Anthropic both IPO, they will get part of every American's and most other countries' pension funds... and when they collapse, a lot of that value will be gone. The initial investors into SpaceX, OAI and Anthropic all will have made their money and put it aside, and we will be left holding bags)
[1] https://www.ign.com/articles/openai-ceo-sam-altman-thinks-it...
[2] https://www.rte.ie/news/business/2025/0511/1512086-doomsday-...
And now this mf'er (or to be more precise, alleged sf'er [1]) goes on the record to say it's bad if we the people have RAM.
Fuck him, fuck everything he stands for.
[1] https://www.theguardian.com/technology/2026/oct/20/judge-dis...
Oh hell no. The hardware vendors want to make as much money as possible, with Micron going as far as to fully exit the consumer market entirely so they can suck as much money out of the AI market as possible.
How that one can fly by antitrust enforcement is beyond me.
[1] https://investors.micron.com/news/press-release/2025/Micron-...
Thiel's comment is also pro-individual-ownership - in context he was talking about the danger that in a mass democracy the people as a whole would vote to have the government restrict the freedom and confiscate the property of successful people (particularly prescient in light of the recent California attempt at a billionaire wealth tax). RAM being expensive is better than RAM being rationed by a democratically-elected government so no one person has too much, as judged by the voting masses.
no, he was lying by omission
"it's quite terrible" only if you ignore all the problems with plans they promote and measure it only in metrics Altman wants to measure, as they support his plans
For start, Altman would charge as much as possible to optimize profits, so very significant part of efficiency increase would become profits of "Open"AI.
As another obvious problem: handing so much power to a specific company has own costs.
There are so many problems here with this specific part - starting from their power dictating how can you use it, access to your data. And that part is getting worse for anyone not planning to obey all decisions coming from USA. We have already seen export controls being invoked.
But wider, if you think that outsourcing thinking to a specific SaaS has no costs, beyond current subscription costs, then I have several thousands of bridges for sale.
Which might be true, I'm willing to accept that, but it does not change the fact his message was incredibly tone deaf.
Cost of living in general is through the roof for years now, many people haven't had the financial ability to meaningfully upgrade their computers for just about the same timeframe - first Bitcoin drove CPU prices through the roof, then GPU prices, once that died down due to ASICs there were a few fads that "mined" RAM, HDD and SSD, and in parallel to the latter Covid hit, disrupting supply chains. After Covid was somewhat over, the Russians invaded Ukraine, then we got the various MENA conflicts after Oct 7th, and now the Iran war is wrecking everything - and ever since AI entered the picture somewhat three years ago, first GPUs and now everything else has exploded in price. Entry level smartphones have all but disappeared. And don't forget our existing computers got less and less useful because first Microsoft pushed the garbage called Windows 11 onto everyone it could and second, everyone and their dog went for Electron resource hog apps.
For Altman to now say "your unused RAM could be of much better use in our servers" is like a giant fat fucking middle finger, telling us to sacrifice even more of our needs just for the opportunity of him and his fellow rich bastard friends to make more money, all while he and his fellow rich bastard friends are executing the setup of a financial crisis that will make 2007 look fucking tame.
> Thiel's comment is also pro-individual-ownership - in context he was talking about the danger that in a mass democracy the people as a whole would vote to have the government restrict the freedom and confiscate the property of successful people (particularly prescient in light of the recent California attempt at a billionaire wealth tax).
Oh for fucks sake. Taxes aren't theft, they provide for the environment that makes these rich fucks able to earn money in the first place. Thiel is worth 38 billion dollars. Even if you'd tax the guy 50% of his wealth, he'd still have more than one single human and their descendants can ever feasibly gamble away.
> RAM being expensive is better than RAM being rationed by a democratically-elected government so no one person has too much, as judged by the voting masses.
At least in a rationed system that allocates supply fairly, ordinary people wouldn't have to run with barely working hardware because they can't afford to buy something new and modern for once as all the new stuff is hoarded by the rich like a LotR dragon hoards his gold pile.
And as evidenced by this week's elections - the people are fed up beyond belief. Either we the people get some relief, the sooner the better, or the situation may indeed devolve into open riots.
if you assume crazy fairy land that is not going to happen, then you can assume cheap and plentiful RAM for better results
Thiel has benefited from wealth being wielded by rich media owners to convince the masses decades long trends in deflation of buying power through inflation of costs are good for them
With such a system giving him a boost he was successful before Millennials and GenZ were adults or even born
Altman came along during ZIRP; being successful when the government is intentionally printing money is not really success
I could see it used as a sort of "sub-haiku" for easy stuff to leave more tokens for the bigger claude models, but I won't be cancelling my Claude sub any time soon.
Coding AIs aren't burgers. Even if a local LLM is as good as Opus 4.5 (when you sink several thousand dollars into hardware ... and I'm skeptical even then) ... the actual difference between the models is huge.
Let's say I use the $200/month Claude (I don't, I use a cheaper plan, but stay with me). If I switch to a Opus 4.5 equivalent local LLM, and it makes even a single production issue that 5.5 wouldn't, that mistake will cost thousands of dollars.
That "cheaper" model isn't really cheaper.
You are assuming that Opus would always be superior; the cloud AI providers can be very variable. They can be very useful for months, but during certain times their quality can drop off a cliff. People online speculate that the cloud providers drop quality during this time when they are training a new model or the service is over-subscribed. You do not have that issue with a local model.
That issue of quality of service is enough many businesses may prefer a locally run model because it will be more consistent.
As is known, LLMs are amazing when verification is cheap, and much less useful when it's expensive.
It actually did. There was an older yacc-based implementation that had a lot of architectural limitations. We decided that it was better to start over with a better architecture, but we did have that older one around to compare generated microcode initially.
There are a few points I'd like to make here.
First, the frontier is jagged. There are errors that Qwen3.8 Flash Next make that Opus 5.5 doesn't make, but the reverese is also true: Opus 5.5 makes errors that Qwen3.8 Flash Next doesn't make. The latter will likely be fewer, but they won't be zero. Outside of really large differences in capabilities, there is no total ordering among models, only a partial one.
As an aside, do you always run Opus 5.5 at max reasoning in order to minimize errors, too? I suggest most people don't, because they have their own intuition for good enough vs. cost.
Second, Opus 5.5's strength are autonomous tasks. For interactive, HITL work, it's not a particularly great choice. This includes its preferring complex Bash commands over tool use and inscrutable Python scripts, but also often not even explaining what it does. It is not a great choice to use as a virtual pair programmer, because following along what it does and maintaining a mental model is harder than with some other models (though probably because of benchmarks that prioritize autonomous work, these and related problems have infected other models, too). You can address that to an extent with CLAUDE.md instructions, but even that has imperfect results.
Finally, if you rely on a model not to make mistakes in order to ensure correctness, your software development process is fundamentally broken. You need to assume that any model will make at least some errors and need a development process that accounts for that. The problem here is not necessarily which model creates fewer errors, but which model creates code where errors are easier to discover. Ironically, more powerful models may create more complex code, making that harder.
You can see an analogous effect with reasoning effort in FrontierBench 2.0 [1]:
"FrontierCode is one of the few coding benchmarks where writing additional (correct) code doesn’t necessarily raise a model’s score. This leads to an interesting phenomenon: models can score lower at higher reasoning effort. Because this non-monotonicity often matches user sentiment but is hard to see in other evals, we consider it a useful feature of the benchmark."
A example:
"This task comes from PostHog’s open-source Android SDK, which apps use to record analytics events. When the device is offline, the SDK writes events to disk and sends them the next time the app starts. The task is to fix a startup bug in which two modules send data at the same time: the live queue sends new events while a legacy integration, PostHogSendCachedEventsIntegration.kt, sends the cached ones. The two race, and events reach the server out of order.
"At low effort, GPT-6 Astra moves cache loading into PostHogQueue.kt, removes the sending logic from the legacy integration, and leaves the SDK’s public API untouched, which produces a passing solution.
"At max effort, the model often deletes PostHogSendCachedEventsIntegration.kt outright and spreads its responsibilities across other legacy files. This also fixes the race and passes most of the hidden behavior tests. But the deleted class is part of the SDK’s public posthog.api surface, so removing it introduces an unnecessary breaking change for every app that references it. Scope blockers, which deterministically check that the model hasn’t modified certain files, catch this change, and the agent fails the task."
What are you talking about, you're fully dependent on an LLM (Opus or whatever) not making a mistake or else you lose thousands of dollars, in production? How is that okayed?
Try it before you go super far into the world of this "must suck".
Most people aren't going full vibe code. If you stear these models, they are almost identical. Seriously. The major score changes in a lot of the absolute latest models are just models doing more things (often when they shouldn't). Having a slightly lower scoring model is actually better in some ways. I digress.
If your model can make a production mistake you are using it wrong. Sorry not sorry.
There are other benefits. Your data is private, it's not sold. Your privacy is kept, that's priceless. No one can degrade your workflows. Uptime is you are online whenever you are online not 3 9s. You also won't get refused asking basic security questions.
Your math doesn't math for a lot of people. It might for you, but I would encourage you to think a little more.
Agreed on that, and actually I'd put it somewhere around opus 4.5 from what I've seen so far.
Good enough may be sufficient for personal projects, but in a competitive landscape where you're using a "good enough" model and your rivals are using a frontier model like Opus 5.5, you will be at a significant disadvantage.
Is this the nerd version of "doomsday prepping"?
So all in all, I put it at Opus 4.8 level, at least. And in some ways it's exceeding 4.8.
Like seat-of-the-pants good as Opus 5.5 but a bit slower and required just a bit more prompting to get what I wanted (which is fine for an app pipeline).
Around 50% of that group says they plan to cancel their Claude subscriptions now that their local or open model setup is performing “good enough.”
If you are just vibing and managing multiple agents and reviewing at a higher level like architecture, you are right.
Open models: we're already seeing this at some scale. https://news.ycombinator.com/item?id=50350004. I do expect to see more of this.
If company X has a better model than company Y, do you use it? Well, that depends: will Y take two years to catch up? If so yeah, you buy X.
But if Y is just as good a few months later, you purchase based on other factors, because you know you can live without the brilliant research X is doing ... since you'll have it soon enough.
That doesn't seem like much of a moat. What happens when a couple researchers go on new parent leave at the same time, or someone's productivity drops for a quarter because a parent passed, or gets divorced, or whatever?
Continually staying 6 months ahead of the distilled open-weight models means you need to keep flipping heads (nobody goes on vacation) until the end of time. The geometric distribution describes that, and it ain't pretty for the expectation value of these companies.
Research does not bring in revenue. Products or services do.
It doesn't cost a company that much to buy a box capable of running the Qwen3.8-Flash-Next model in the GP comment and never pay Anthropic again. The cost will be even more reasonable once the supply issue is solved for that hardware.
AI-as-a-service is a unique product because when the interface is natural language, there is no lock in. VS code has a hotkey to swap between claude, openai and a local model. Compare that to the difficulty of swapping from macOS to Windows.
I don’t have any stake in this and I hate the idea of concentrated wealth and power, wherever it is: government or private businesses. But would I bet against it? Probably not.
We're the only people that can spend this money is not a moat.
If you need to buy staggering amounts of compute, spend staggering amounts in utility costs and lobbying, and keep spending that as your hardware amortizes and your operations go on, what you have isn't a moat, it's a standing army. It's extremely expensive, and the day you start having issues, mercenaries will cut your throat before you know it.
But the fact that were even talking about Anthropic like this only a few short years later demonstrates that that was only kind of a kind of moat.
Check out mythic.ai for one example.
Banks are going to want that "capital" to materialize into gains at some point. The water in that moat is evaporating.
Hardware's ridiculously expensive. Then, there's no good place to put it, because the new breed of tech bro managed to propose the data center buildout in the most exploitative, alienating, and electorate-angering way possible. So compute's hard to buy.
I guess maybe another anecdotal comment, but I'm an engineer myself, have a home server for simple stuff, and I pay $200 a month to chatGPT for a pretty simple usage and really don't see myself go and run my own open source model, especially at the speed at which they evolve and change.
99% of the world will not setup their own open source model, just like 99% of the world is not setting up their own servier with ssh copy whatever the dropbox comment was saying.
If it were not for the x86 architecture they would almost have models that could answer those queries on their desks. (Gemma 4 E4B QAT is surprisingly good at this stuff; Gemma 4 12B QAT is great at it).
If I was Sam Altman I would consider giving Intel and AMD money to not develop unified memory architectures.
Opus for planning and the very difficult parts, Qwen for the mundane ones. It still involves a lot of manual model switching (but opencode and Pi make that relatively easy).
Strix halo is double the cost because of...
Do you mind sharing more details about your setup? I have been using opencode with lemonade for inference on my strix halo but haven't quite hit the sweet spot.
Here's the gufo command:
gufo serve 1lm \ -m /mnt/LLMs/models/unsloth/Qwen3.8-Flash-Next-GGUF/UD-104_XS/Qwen3.8-Flash-Next-UD-1Q4_XS-80081-0f-00083.gguf \ --served-model-name Quen3.8-Flash-Next-UD-104_XS \ --host 127.8.8.1 --port 10083 \ -c 262144 -n 16384 -j 1 \ --prefill-chunk 8192 \ --speculative mtp \ --mtp-model /mnt/LLMs/models/unsloth/Qwen3.8-Flash-Next-GGUF/MTP/mtp-Qwen3.8-Flash-Next-shared-Q8_0.gguf \ --cache-disk ~/.cache/gufo-fork \ --cache-disk-bytes 68719476736 \ --cache-disk-staging-bytes 8589934592 \ --cache-ran-bytes 4294967296 \ --log-progress
I feel like the industry hasn't solved that alignment problem yet... We have no way to verify the entire models landscape of intent.
Paid models have an incentive not to do that because if the market discovers it there is a huge economy penalty, but the free open source models don't have that disincentive.
False.
I worry more about Claude.
For those keeping score, this probably isnt exhaustive- but here's some greatest hits;
#55743: Claude Max account email auto-injected into LLM context with no opt-out - May 3, 2026 https://github.com/anthropics/claude-code/issues/55743
#72518: Anthropic embedded spyware in Claude Code - June 30, 2026
https://github.com/anthropics/claude-code/issues/72518
#21151: No indication of WHICH file for READ tool - January 27, 2026
https://github.com/anthropics/claude-code/issues/21151
#42796: Claude Code is unusable for complex engineering tasks with the Feb updates - April 2, 2026 https://github.com/anthropics/claude-code/issues/42796
#73408: AFK auto-continue on unanswered AskUserQuestion is undocumented and has no opt-out - July 2, 2026 https://github.com/anthropics/claude-code/issues/73408
(Also worth a mention, OpenCode PR #18186: anthropic legal requests - March 19, 2026 https://github.com/anomalyco/opencode/pull/18186)
What exactly are you claiming is false about my statement?
The list you presented are mistakes Anthropic made or cases when they put their own interests ahead of users, but I don't see how it disproves the point I was making. None of those are examples of embedding backdoors or stealing secrets to take over or sabotage a system. They are categorically different than what I'm asking about.
This just fundamentally isn't true. They run on the same economic model, but they have to release the weights to gain credibility, unlike the American labs.
> Do you ever worry whether these open source models have been trained to embed backdoors or leak secrets in very specific circumstances at some point in the future?
Open weight models are extensively analyzed. It is extremely unlikely there is an intentional 'backdoor when X' concept added to any of them. Not impossible, but unlikely for the same economic incentives you see in the other labs, and given that no evidence of anything along those lines has surfaced yet.
In my opinion, to suggest open-weight models are extensively analyzed while being net new and continuously released and the model itself being unintelligible to humans seems a bit premature.
It's possible that closed-weight models end up following the pattern of being arguably less secure since there are fewer eyes on them analogous to open source software but I don't think there's a perfect match here.
... as long as the reviewer model hasn't been deliberately lobotomized, that is.
(I don't disagree with the [dead] comment, either -- fortunately I tend to make small changes to code I'm very familiar with, so it's fairly easy to vet in person. The more code gets generated by the clanker, the more nervous I get, just because it's impossible to keep some complacency from creeping in.)
Auditing output by contrast is a continuous cost. On the third hand, it's one that most will want to pay anyway.
The "embed into shipped code so subtly humans can't spot it" vector is sci fi ish. Less plausible than old fashioned trusting-trust issues, and most folks have managed to sleep at night for the last 40ish years that spectre has haunted.
[1] I guess the model could return output that buffer overflows the harness and exceeds its intended ability, but that seems far-fetched too
just read the damn code yourself!
It's basically impossible to do this. Claude produces so much excessive / verbose code and comments that reading it would take far longer than simply doing the implementation "by hand". It's kind of like, Claude can write the code in 1/2 a day. It can be reviewed by another LLM, unit tested by another for 1 day. Or I could write the code in probably 3-4 days. Reading it and validating / testing it myself though would easily take 6+ days.
Now if I were fully in control I wouldn't run code that wasn't truly reviewed at all, but in the situations where Claude is being used that's not my call. TBH a lot gets caught, and I have some automated scripts that find a lot of the more problematic stuff automatically, flag it, send it back and make note for the reviewer model as well.
I will also spot check areas where I think an issue could arise.
And my god, $15 of extra credits ended up giving me only 30 minutes of Claude Code before running out.
The subscription prices can not be financially sustainable. There is no way that is remotely close to the actual cost of running.
The median user is using AI as a search engine, meme creator, and/or chatbot… at best.
The additional cost created by the subset of programmers using subscriptions rather than tokens/api can easily be categorized as a marketing expense.
Remember that I referred to median user here.
> Do you have any data you're basing that claim on, or are you just speculating based off of you projecting?
Anecdata? I guess it depends on what you count as “data”.
I am currently training a lot of people directly or indirectly (train the trainer type of stuff) on how to use AI professionally. I always start with “what do you currently use AI for?”. The answers I listed above are largely the categories that people have told me and the folks I train, with the exceptions being a handful of non-technical folks who have leaned into AI and have used agents and/or created apps of some sort.
To add some categories that I didn’t initially mention, some folks also use AI for drafting/writing documents and for creating meeting transcripts.
> AI maximalists would have you believe that people are using these things as therapists and having conversations with it that are more intimate than any lovers'.
Chatbot… this is something I hear, but I’m guessing I don’t hear all of the examples since it’s personal rather than professional. But some folks certainly say something like “I treat AI like a non-judgmental friend that I can ask personal questions to”.
> AI minimalists would have you believe that no one is using these thing except at gunpoint, and even then.
In my experience, this is almost exclusively the domain of old/older people who are happy to let tech pass them by and young people who are virtue signaling to their peers (or at least trying to).
> Unless you have some reason for us to believe you that isn't just "trust me bro", your portrayl of the median user is as suspect as any other.
It seems like you’re new to HN. I understand your skepticism. There is a relatively high density of actual professionals here. Trust me on that, bro.
My data set numbers are in the 100s, and it is not a statistically representative cross-section of the US or world market. That said, I’m guessing that I have a fairly good range of data points for the US market.
> or are you just speculating based off of you projecting?
This is also a strange thing to say. Why would you assume speculation or jump to projection as the basis for it? Nothing in the comment that you are replying to hints at projection.
In the S-1 filing Q3 was profitable even with training included, if you exclude stock based compensation.
Q2 was profitable on all operating costs (incl. hardware, employees, training, inference). They were not profitable on non operating costs though. That’s partially interest, but mostly, when their valuation goes up, they’ll have to issue more stock (absolute value) to early investors after IPO. (so, them going up in value is a loss, as the shares they’ll owe will be worth more)
Their profitability will likely crator as they start building their own data centres after the IPO, so that’s the best results they’ll have in a while.
Anthropic is super opaque with what inference costs at any given moment on the subscription and cache is just super inconsistent.
That is like saying selling hamburgers is profitable if you exclude the cost of beef.
If under the scenario you outline where they don't need to upgrade then AI is commoditized and margins are razor thin. And that is fine. Great even! But it doesn't justify Anthropic's trillion+ dollar valuation.
Anthropic is running 85% gross margins on inference.
The API Gross margin is even higher.
The purpose of the high API prices is, in addition to trying to get more money, to anchor the high price, and then make the subscriptions seem like a great deal. But inference is just not that expensive, not when provided at scale and with huge batches.
You can see this with the various Chinese providers doing inference on open models, which are unlikely to cost less to run than the models at the frontier AI labs. A single 8-GPU GB300 cluster doing high-throughput batch inference on Kimi K3.1 Pro, which is the largest open weights model available, can do something like 300 000 tokens per second, that is, 10T tokens per year. That cluster, with current insane inflated prices, costs something like $1M to buy, and $50k to run per year in electricity. At $10/Mtok, that pays itself back in 4 days. Do you think that that's a reasonable price?
The subscriptions were probably priced with reasonable positive margins.
The huge expenses are in training and pushing the models forwards, with the gigantic RLAIF runs we know they are doing. Then once you got the model, you can scale up inference at low cost and want to serve as many people as possible.
You wanna know the real price? It's easy, just check what AWS, GCP or Azure charge for tokens.
This excludes the R&D costs, including training, where most of their costs lie.
The basic truth of LLM economics is that inference is cheap. The NRE cost to get where you can get premium prices for inference is sky-high. This is part of why the AI companies want to drive the scale up so much.
My guess is a mixture of both... There's a reason why enterprise customers have to use API based billing.
(We also had a stint of using Grok 4.7 and that was the same price for our usage. Opus 5.5 seemed smarter though. After taking into account caching both average around $0.60 per 1M tokens)
I've been running into the same thing, and adding separate OpenAI seats for employees rather than paying obscene overages to Anthropic. Thankfully Opus has decreased usage quite a bit, and made this less of an issue as well.
It's completely unclear to me why this is the conclusion everyone jumps to instead of "holy cow they are milking it on API pricing."
I mean, have you ever run a local model? How much electricity did it cost you to get 1,000 tokens out of it? Like a fairy-dust of a cent?
The AI labs' cost is CapEx. Marginal cost of tokens is zero.
They lure you in to do "automatic top up", which gets quite expensive if you dont watch it :-)
Welcome to... any subscription system ever? If every person with a gym subscription actually went there say three times a week, or every person living in a dorm paying flat rate for electricity started mining bitcoin 24/7, these systems would collapse. That doesn't mean it's not sustainable, it just is how subscription pricing works, by having some people use its offers a ton and others barely at all
And there's a hard place that everyone freaks out at API prices and on-demand Opex models outside of a niche (I use API for personal research for quite focused queries and slightly more reassurance on being the customer and not the product) but this is very competitive + not sticky.
Overall it would go through weekly usage within 40 minutes, if not for 5h session limit.
Running on 10000 computers will just hit all the limits at once.
they're setting a precedent that artificial intelligence should be free, or at least at-cost (free from profit), but robot labor (API usage) should not. I am all in favor of this dynamic. They built / maintain the robots, so they profit from their labors.
Even if they are a loss leader, they serve a valuable purpose: people who get used to using AI a lot via the subscription then lobby their companies to start using it. IMHO this is one of the reasons Anthropic was able to make major headway into enterprise.
The risk for investors is they don't succeed in getting regulatory capture because without that they don't have a long term business model
If we believe all the hype about programming efficiency, then regulatory capture in the US becomes just one more in a list of many steps the current administration has taken to chase software development jobs overseas.
The current SCOTUS majority will change any ruling as needed to satisfy the wishes of the oligarchs.
The market anticipates the unlocking; when the shares become public, buyers swoop in, so the stock decline is much less than expected, if at all. There is no new information here. If everyone is expecting shares to unlock, then it’s priced into the market accordingly.
2) Valuation aside, SpaceX has a massive moat. It’s crazy hard to do what they’re doing. Anthropic increasingly has zero moat, with the market rapidly coalescing around commodity open models.
"SpaceX (ticker: SPCX) was priced at $135 per share during its initial public offering on the Nasdaq on June 12, 2026."
current price is $134 , give or take a dollar.
1: https://www.forbes.com/sites/tylerroush/2026/06/12/spacex-op...
They’ve now got all their eggs in a new basket. A basket that isn’t entirely in working order yet, and whose market viability is still uncertain because it remains to be seen whether a market that’s used to panniers really wants a laundry hamper.
$/kg to orbit is not as important a deciding factor as many people assume it is. It’s typically a small portion of overall costs, and there are often cases where a launch provider that’s less expensive in terms of $/kg to orbit is actually more expensive in terms of total cost.
Starship is going to be a particularly weird one to adapt to. The “deploy out the side” design may make it straight up incompatible with a large subset of spacecraft. For those where it is workable, migrating commercial payloads to it may still require the development of new satellite buses, which is something SpaceX doesn’t seem to be working on for the benefit of anyone other than Starlink. And such a large rocket means most potential customers will have to plan on ridesharing with even more co-passengers, which means even more room for scheduling conflicts, compromises on initial orbit (and therefore cost to reach destination orbit), etc.
That’s just assuming Starship development goes smoothly from here on out. Which seems optimistic considering that, even by the standards of spaceflight, its development thus far arguably qualifies as cursed. Considering the gap between original design goals and reality is already historically large, I certainly wouldn’t be taking any cost estimates at face value right now. Like for any launch vehicle program, we won’t know its true capabilities and costs until it’s operational.
That's the primary angle of depreciation of Anthropic's outlook, is that as interest rates are running hot, they need to grow ever faster to justify the $2B target, at the same moment that they're growing slower.
So many people who don't actually read and evaluate things but just buy stocks based on vibes and what they see on Twitter. Which is to say, the same people that buy and lose money on most IPOs.
Arguments are being made that, with the midterms over and 2028 just around the corner, a less sympathetic administration would create significant headwinds, adjusting the valuation of Anthropic and OpenAI down.
What we don't know are what Anthropic and OpenAI have planned and in the works.
A large population of extremely skilled and talented people are being laid off by their employers to make space for AI investments. It's likely a significant number of them will start their own business with AI.
AI is also gaining a much wider audience and acceptance. It's likely over the next few years, AI demand will grow, not stagnate.
That said, IPOs are usually liquidity events for LPs and for something really disruptive and unpredictable as AI, I won't be engaging in their IPOs. Only time will tell if this was a good decision, or bad.
I am highly confident that Anthropic will 100X their token sales volume. I am also confident that token prices will continue dropping through the floor due to competition. So to 100X their revenue they will have to 1000X or 10000X their token sales. That seems a stretch.
Using the models for data processing (a lot of input/output tokens) and for role-playing will use more tokens and thus cost more. However, those use cases are not common when comparing the percentage of the user base.
The money maker is agentic coding where people get the models to perform long-running tasks. This is why Open AI and Anthropic have been pushing these hard. However, there is pushback on this, especially among businesses looking for value for money/cut costs [2] like Apple [3], [4].
[1] https://aws.amazon.com/bedrock/pricing/
[2] https://www.ibm.com/think/insights/tokenmaxxing-dead-long-li...
[3] https://www.bloomberg.com/news/articles/2026-11-04/apple-cut...
[4] https://9to5mac.com/2026/11/04/report-apple-reduces-employee...
The issue is, at some point for AI to grow more in usage it will need to decrease in cost.
We are already seeing this with some companies reducing AI usage or moving to open models due to high costs.
AI will surely boom there's no doubt on that, but it won't necessarily produce the revenue that Anthropic investors hope for.
many companies are now using both, it's only been a few months since it really picked up
Do we know if current pricing is sustainable?
Something like 99% of the people who have interacted with an LLM chatbot lose interest after a month.
100% of people who have to interact with an LLM chatbot taking the place of a human customer sales or service agent jump to a competitor.
Outside of the tech bubble, the LLM hype is already over.
This is not remotely true. It literally couldn't be true, considering Open AI have 1.2 Billion weekly active ChatGPT users.
I care about whether it's going to grow in value/price from that point on or not.
And it's hard for me to imagine it growing even higher than the current valuation.
However this effect could just serve to make those who own the compute ridiculous money, while crushing those who need to lease compute.
I now do like 90% of my routine software dev/sysadmin stuff with Luna on Medium by default, some proportion of which could itself be offloaded to a small Qwen model.
Indeed, 2026 was basically that: the LLM providers increased token costs and companies cut LLM spend to the bone. That is why Anthropic's IPO is struggling and MS and Google have failed to meet AI-related revenue targets for 3 straight years.
What's you're point there? People laid off because of AI will immediately shift into AI-centric small business, and start hyping AI?
That doesn't make sense to me. People being laid off are likely to be bitter about the cause of their layoff, and even if they all start their own businesses (which is unlikely), most of those will fail, leading to more bitterness.
If these giant monopolies couldn't squeeze revenue out of these employees using AI, why would we think that they will be able to do it by themselves?
The vast majority of them if they want to be entrepreneurs will be better off starting a new lawn care business, or a bicycle repair shop. They can use AI to help them run that. Getting paid $175K per to write ruby CRUD was a moment that is over, hope you didn't spend all your salary on toys and rent.
> If these giant monopolies couldn't squeeze revenue out of these employees using AI, why would we think that they will be able to do it by themselves?
You assume they are competent. In my experience a monopoly's profit margins makes it very easy for incompetence to flourish.
> The vast majority of them if they want to be entrepreneurs will be better off starting a new lawn care business, or a bicycle repair shop.
Really? Experience engineers are better off starting a lawn care business or a bicycle repair shop? I do not believe this for a second. There are many areas engineering skills can be used in.
Outside of programming and a couple of other sectors I’m not sure that is true. It isn’t visible in the unemployment figures.
The unemployment figures is mostly a political projection, like CPI.
In 2026 everything fron sewage to chickens to oil to education is now tied to AI. The entire economy is now, for good or bad tied to AI.
What most people either didn't notice or understand is that the U.S. was entering a recession and would have been in one if it hadn't been AI
Tied how? How strong is the corellation between these? I'd say very weak, other than everything having to compete with AI for resources.
The same is true of any technology tied to powerful personal computers or the internet - people built and designed beatiful houses, great cars, airliners, wrote books and ran restaurants without computers or the internet.
It has even become a meme how putting smart features and lcd screens into things like washing machines have made them strictly worse.
Is it maybe more accurate to say that the US is in a recession, outside of the industries that are fueling AI?
Life's pretty recession-y for everyone else imo
Even if it takes them 3-6 Month to open up a new field, its going to happen steadily now and were it started, the AI investments will be recuped by layoffs.
Artists -> Translators -> Journalists -> Software engineers -> mathematicians -> lawyers -> Accountants -> ...
Its already disruptive
We're not paid just for what we do, we get paid because we accept liability for our mistakes. Professional liability insurance is a large part of why licensed professionals charge so much money for their services.
If a lawyer needs less people preparing everything, the industry will also shrink.
Lawyer as SaaS companies will just introduce some insurance and will clean up from the bottom until AI can do all of it.
We will see if we will have one expert ai lawyer per topic or type of lawsuite or if they will get generic good enough that it will not matter.
Even people defending or suing by (successful) themselves became a thing already last year.
Lawyers already need less people to prepare things. That revolution happened several decades ago when computers and word processors became normal.
Lawyers as SaaS companies has been a thing for 2 decades, and remains a tiny niche of the market because it's not the kind of lawyering most clients need. A lawyer SaaS is basically just the equivalent of an underpaid, underresourced public defender and will yield about the same quality of work and results.
We will see if we will have one expert ai lawyer per topic or type of lawsuite or if they will get generic good enough that it will not matter
LLMs are only as good as their training set. In order to be a topic expert, they'd have to be retrained at least every week to account for changes in statutory law and case law. It would be trivial for a human lawyer to outmanuever an outdated LLM when the laws changed...and in the real world this happens all the time between human lawyers when one side hasn't done its research on changes to the applicable laws.
Even people defending or suing by (successful) themselves became a thing already last year.
People successfully defending or suing without a lawyer has been a thing for over a century. It's fewer than 0.1% of legal proceedings. Most people don't handle their own legal cases for the same reason most people hire a plumber, or electrician, or order DoorDash instead of cooking their own dinner. Also...almost all of the people using LLMs in the past 2 years lost their cases, because their LLMs hallucinated laws or facts and they didn't know enough to catch it. Even lawyers dumb enough to use LLMs to draft their documents got caught because they were too lazy to review the output. Courts were willing to forgive those mistakes a few years ago before it was common knowledge that LLMs make things up, but now they're no longer willing to give the benefit of the doubt.
Whatever we had before on full text search, LLMs are a lot better. And i'm not just feeling this, i read plenty of articles about exactly this.
A LLM doesn't need to be retrained every week, it needs to have a base corpus of the field and strategies on how to find current text. They will load it into context and argue.
Then another LLM will take this and formats it for the court.
We will see but i think you overestimate how much automatisation and co current lawyers have and use.
Mine already has been.
At best some layers will work faster, but automating their work or people not requiring them? Usually when you contact one there's a lot at stake. Same with accountants.
Software engineering and translators is a different story. Most of the times if something goes wrong nobody cares.
The avg Lawyers might do more complex but again a lot of 'normal' things. Day to day things the next SaaS AI company wants to 'fix' and they might also accept basic liability.
Then you have the 'tv lawyers' the ones who do non normal things, they have A LOT of apprentices and young lawyers going through papers, research etc.
They will for sure have the expert as a human, but that person will need a lot less support.
You shouldn't need an accountant today already. If you have an digital order and a digital bill and a digital signed delivery note, it can be booked automatically. AI will find risks, summarize them etc.
AI is already helping here closing the gap with better OCR etc.
I really think, that when big companies introduce the agentic layer to their products, there will be less people needed and if you need people, they get less paid because they will need to understand less.
Wouldn't they tell us before IPO? Unless it's not that promising...
> AI is also gaining a much wider audience and acceptance. It's likely over the next few years, AI demand will grow, not stagnate.
That does not mean the current players profit from this and the infrastructure investments were necessary at that scale. There is competition, there seem to be enormous efficiency gains and it's questionable that demand for even bigger models is large enough.
If datacenter production sees significant headwind then it’s very bad for ANT. Securing compute is listed as one of their major risks to the SEC.
That said, I'm also somewhat skeptical that there is even much room for growth left unless somebody finds a new, important, and token-hungry use case.
They are also (like other models we know about) experimenting with the architecture. [2]
[1] https://arxiv.org/abs/2501.12948
[2] https://sebastianraschka.com/blog/2026/kimi-k3-architecture-...
[1] I don't, I think China is primarily limited by compute and will be at least for the next couple of years, probably longer.
If Anthropic manages to start producing new pharmaceuticals with their internal models or something that thesis could make sense but not if 90% of their revenue comes from enterprise API sales which would necessarily have to be accessible to the public. Which are then inevitably fast followed by Chinese models 3 months later.
Your Railroads and the Internet parallels are so very apt. We certainly are very productive even in 2026 off those but almost no one remembers how many large behemoths were wiped out in the U.S. because of them.
Yes, it does seem to me that the AI of 2026 will be like the Railroads and the Internet - fundamentally transformative technology that changes how things are done into the future but there will be blood in the short term, which is why I would not be participating in the IPOs directly
It seems like much of the tulip mania has become the stuff of legend.
From what I have read, it seems closer to baseball card trading than AI or dot com.
You can find baseball and pokemon cards trading for thousands of dollars right now on ebay.
This would seem insane to a culture just valueing the cards as pieces of cardboard.
It would be like hundred of years from now people viewing our society as wanting to buy Mickey Mantle cards instead of houses. It is a very small number of people involved with this. It is not a widespread movement to hold your retirement money in Mickey Mantle cards.
In an interesting way it shows even better though he we are susceptible to these type of mythologies. People don't even question the absurd givens of the narratives of the tulip mania. We just repeat these narratives to each other as if they are historic fact.
See https://en.wikipedia.org/wiki/Railway_Mania and https://en.wikipedia.org/wiki/Pets.com for some info on the other hype cycles I mentioned.
You really don't know what we have here, do you? Let me illustrate what you can do with Qwen3.8@q4 locally with about 2TB of storage and a small cloud server to host a website.
Locally you scrape and process every elected official and start tracking what they're doing, what they're promising, and what they're not doing. You build embeddings and make them all geotagged in a database and just start with the largest cities.
The website provides public facing search. User drops in a zipcode, city, state, specific rep's name and gets what's relevant to their results.
All of this is to say, we literally have the technology to hold accountable our representatives and then show everyone objectively who's doing good and bad for their constituents.
It wont take much to solve the problem of bad representation after these last two years. The hard part is literally making sure people going to a polling station can immediately know who the heck the 10 - 30 different people are they're voting for, and track them efficiently within their own specific parameters.
Regulation is coming for more than just the inference providers.
Not making a leader board to be clear though. The goal is to build a branchable leaflet style page based on zip codes that lists ALL the people that were elected. Clicking into the each person will yield details on any legislative actions, their attendance, publicly made promises, etc.
I'm really sick of hearing lip service from politicians so I want to build receipts that even a child could understand.
I find myself wondering what "less sympathetic" and "significant headwinds" mean in the minds of people in the upper echelons of the AI industry.
Rules for training frontier models? Requirements for renewables to power data centers? Actually having to pay people for the content you train on? Data center rules that would end them going behind the backs of local electorates to get governments to approve the building of data centers?
Agree. these valuations will hold. Pundits made the same prediction after Facebook's IPO in 2012 that everything would soon crash. here we are...
Lets set aside speed. The results are ALWAYS terrible compared to CLaude Code.
Sure I only have a 12GB GPU - is that what it is? If I had a 32GB CPU would the experience be as amazing as Claude Code is?
Why is my experience of local LLMs so average?
You can maybe replicate Opus 4.5 to 4.7 if you have some moderately expensive hardware. And that is definitely good enough for many people.
The smallest model that can even pretend to be a decent coding model is Qwen3.8 27B, which you realistically need 32GB of VRAM to run without breaking or limiting it in some really noticeable way. This gets you something like a slightly less capable Opus 4.5 with a much better work ethic.
They seem to be happy with their small usages.
For everything I do, i see huge differences in Quality and Speed.
Even when I try an open model on a 4090, the PC gets warm/hot very fast, I struggle with context length.
Having currently a very tech-intensive chat running for more than one week, this is where the frontier models really shine: Its no more 2024 or so, where they lost context on large discussions, you can do really great work with the frontier models - the local models will be a thing in a few years, if RAM & CPU developed further, this will happen: in 2035 like, you will be able to run "Claude-like experience" on your iPhone.
I write 5-20 paragraphs of things I want and it goes away and just builds it.
Something like this (just an example):
"Build an application with a webserver in rust. It has 4 stages, first one is an input module... second one is refinement ... third one is an image detection with current fitting ml model ... last one is finetuning."
Then i would describe every basic page with a few more paragraphs.
This would work for half an hour or an hour, consume 100-400k of context tokens and then I would test it, write down 10-50 small sentences (every sentence either fixing a bug, or making it better) which tehn takes again like 20 minutes and i'm nearly done with a complex enough tool providing real value and actually working.
A big kimi, which is hard to run, might be getting close to this but we do not talk about running it on 1-2 5090.
The other thing is, thats more like a practical experience, even if you use them through Open Router (for testing and co), the initial results of something like Opus 5.5 are ALWAYS better. More details, better quality, better understanding of usability of the tool.
And yes Sonnet 5.5 is crazy good
> In a work tree build a feature where each column cog action menu has a "Column statistics" item - clicking that item opens a modal dialog which displays statistics about that column - for text columns that's most common values, stats about value length - for numeric columns it's more interesting than that. Let's do a bit of percentile analysis in there too. Under the hood it works by firing off SQL queries using SQLite window functions and suchlike. For the moment let's have that option only available if SQL queries are allowed against the database, and the SQL queries are built in JavaScript.
You can see a screenshot of what Opus 5.5 built here: https://github.com/simonw/datasette/pull/3038
> They seem to be happy with their small usages.
Oh yes, we're terribly inferior people. Just living on scraps from the masters' tables.
I do plenty of experiments of things i always wanted to do and they just don't work with local models.
Finetuning of some ML models for a specific usecase, some 4k demo idea i had, small browser games to show others game mechanic ideas i had for ages, analysis tools etc.
Even just playing around with some Ghidra eats context like nothing.
If that were really the case, basically nobody would spend so much money using them.
You should factor in that undeniably most people don't have the means to run even a 27B dense model locally, and therefore do not actually know what they are and are not capable of in that context.
And you might want to look at openrouter's rankings.
If you are comfortable automatically assuming that "doesn't use Claude or Codex" is equivalent to "not a good engineer" or "bad engineer", then I think you are probably lost to more nuanced discussions.
It still can't compare to Opus 5.5 max in terms of speed, capability, and stability.
And when I calculate all the money I've invested, directly subscribing to the service would actually be cheaper, more efficient, and more stable.
Open source is great, but don't glorify it.
The architecture of the code I develop has never been simple. If you're someone who doesn't even work on things that need to be deployed to something like k8s, then basically, open-source models are indeed enough for you.
the original comment mentioned only 16GB of VRAM.
You know talk about Open Router and 27B dense model?
We are not comparing an even more expensive Hardwaresetup to run local models with proper context window and parameters.
Can i run some open model on a super expensive Mac Book or co with some kind of okayisch results? Yeah sure but you know what the t/s will be and yeah Frontier is still Frontier and Kimi, GLM, DeepSeek are not Frontier. They lag half a year.
You are not getting Opus 5 or Sonnet at home for now.
Why are you so offended by this?
- 2023 using Cursor
- 2024 using Cursor with Agent mode
- 2025 using Claude Code
- 2026 using Codex with Astra 6
I've try every new model when it comes out from both OpenAI and Anthropic. I cannot justify leaving Codex and Astra 6. The token burn on Anthropic is always astronomical for me. Codex just works, and I see no reason to move away from it.
There is, for sure, a limiting factor in the x86 architecture, but I think the RTX Spark will change that a bit, and maybe AMD's subsequent Ryzen AI/Strix Halo chips.
In any case, 128-256GB unified memory systems running MOE models is the "cheap" option, RTX6000 is the step above it.
I'm not familiar with these. How much does one cost, and from what store / what product name?
I am digging in on Pi (in lovely Paseo and exploring Pendant when in VS Code).
What would I find better or worse in Kilo?
The newer/larger models are good though, not as good as Claude, but not every task requires a model of that level. Maybe yours do, but for simple coding tasks, summarising text, etc a small local model runs just fine, and is a lot cheaper than Claude
You can try that model (and many others) using a cloud host first, and then only if you like it do you need to buy something to run it.
RTX 5070 Ti and RTX 5080 both have 16 GB.
RTX 5090 has 32 GB.
Good luck finding any of them for reasonable prices, though. I paid $3,000 for my 5090 early last year. I winced when I bought it then. Today, it's like $7,000, and I wouldn't even consider it at that price.
But I'm mostly a gamer that also does local AI as a hobby. If the RTX 5090 was $7,000, I'd still be on my RTX 3080 and still just not started doing local AI and have to turn down the graphics settings in Cyberpunk 2077.
You can surely build an LLM server with an R9700 and 32GB of mainboard RAM for less than the 64GB M5 Max Apple Studio, which is probably the next most competitive model up. It'll be better than a Strix Halo but less flexible.
- 4090: Going for $7500
- 64GB M5 Max: Starts at $3,800
- R9700: $2700
Baseline of $20 or $100/month Claude or OpenAI subscriptions. Alternative posed of running models locally. A guy says he did it and results are lackluster on a 12GB video card. A number of people post that that card is insufficient; use 32GB+ cards. I look up prices of 32GB cards, and they are all very expensive. So from a practical perspective, people are suggesting you spend thousands of dollars on a video card to get results that are almost as good as the subscription plan they are on. It doesn't add up, unless one of my assumptions is incorrect.
Having a 4090 at home running agents all day is not a $20pcm level of usage, so take the $100 plan. Looking at the S-1 filing, today they are not even close to breakeven. It is reasonable to assume that will become a $300-400 pcm plan.
The launch MSRP of a 4090 was $1600. There has been significant inflation since then, so say $2000. Eventually supply will meet demand and you'll be able to buy one for that price again, or lower due to the economies of scale they must have now.
So, in 2-3 years time once it has all settled, you pay off your 4090 by saving 5-6 months of Claude usage. Maybe the local model isn't good enough for complete replacement, but all it needs to do is drop you to the $20 plan to be worth it. I think this is a large part of why investors are getting jittery now.
Right now you could extensively use an M5 Max for a year and sell it for almost exactly the same amount at the end. Maybe even more, at the current rate. And it can do other things for you: host your VMs, edit high def video, whatever. Even if it stops being useful to serve LLMs, it might be useful to run as a server for other tasks for a whole decade. It will be a fast, supported Mac for seven years. For the R9700, maybe gaming?
Depends what you want, doesn't it? Or are we all expected to conform?
For me personally, right now I have a secondhand M1 Max macbook pro, and that is my local AI machine, because my needs are exploratory and I want to run it here. I want to learn by comparing models. I want to learn how to run them and advise others how to (local inference is not a zero-sized market).
My next step up will be to take the apps I'm looking at building that run against these small models, and host those apps online with some sort of cloud inference — openrouter, digitalocean, cloudflare — for those same models.
If I ever find I want to vaguely describe some desktop application before bed and come down the next morning to an aggravating slop version of it written in Rust, I'll consider Claude.
For a local non-laptop, I've looked at the M5 Max and the R9700. I've also looked at the M5 Pro, which in practice is about as fast at token generation as my M1 Max (because the M1 Max's memory bandwidth is strictly theoretical), and smokes it at prefill. That alone might be enough for a "leave it running" mini AI box, and it will be a nice desktop machine. I just don't need it enough yet.
As an aside there's this thing — somewhat interesting, though constrained:
Uses the 5090 mobile chip, which is pretty good speedwise (about the same as the R9700) but 24GB. Small, tidy, on-prem. I think they have missed the boat on pricing, but if they can ship in volume, maybe not.
3D games give me motion sickness so I don't have powerful PCs anymore. In fact right now I don't think I have a desktop/server PC that can power up at all, and that alone makes me wonder about building an R9700 machine.
However even if you had a private data center you wouldn't be able to beat Opus 5.5, Astra or Gemini 4. The open weights models are a tier below these three.
Good enough? Quite probably, unless you're in the "coding is solved, specify loops and walk away" frame of mind.
32GB (or say a 48GB-unified RAM unified machine) gets you Qwen 3.8 27B at a good adaptive Q4, maybe Q5. Muse Glimmer at Q4 and Q5, etc.
These models can code well enough in an agentic way, I find, and the chat-to-assistant scenario is just fine in Unsloth Desktop or LM Studio.
A slightly lighter harness helps; a lighter system prompt is entirely possible, especially for Muse Glimmer, which is post-trained so effectively for agentic helper tasks that it barely seems to need a system prompt at all except for a "plan" mode.
Tool use is all fine, all that.
which is where the entire software industry is going
But will you want to do that when your tokens are metered at non-subsidised prices, and yet there is absolutely no clarity (ever!) about token efficiency, how the model decides to spend your money, how it might ever decide to stop wasting your money and get to the point, etc.?
The specify-loops-and-walk-away model absolutely points to local/on-premises/private cloud AI.
Because otherwise you're letting Anthropic or OpenAI decide how to spend your money. And they are going to be biased towards spending more of it.
I don't think Qwen 3.8 is quite there yet with the loops stuff. Definitely seems to be better at deeper and deeper one-shot puzzles.
"SemiAnalysis, a research and analysis company specializing in the semiconductor and AI industries, found that a $200/month subscription plan is far more generous than previously thought. According to its calculations, while it's widely believed that a $200/month plan maxes out at roughly $2,000/month in tokens (assuming API pricing), the actual numbers are closer to $14,000 for ChatGPT and $8,000 for Claude. "
https://cybernews.com/ai-news/chatgpt-claude-usage-costs-ope...
This will change in the future.
Then you would go try it and be back using windows to game a week later.
Everything considered, linux is good and has decent capabilities including gaming and whatever you can throw at it but only do oranges to oranges comparison. Games and apps built for each os works fine on the respective os, well atleast it is true for mac and linux. Windows is very unreliable.
Your experience is average because it _is_ supposed to be average. The people using local models tradeoff taking the moral/technological high ground vs actually using something high quality to get a job done quickly and with high quality, and largely autonomously. They are ok with subpar output because they will handhold the model more, or the (subpar relative to the frontier) outputs are sufficient for their toy use case while being cheap and accessible, with the added benefit of being able to run it on their private data without sending it to a 3rd party.
Almost no one at any job where software is the main economic output is using a local model to code. If you actually require intelligence to get a hard job done, you should just use a frontier model.
Or, you know, your own intelligence.
What I do not understand as an engineer is how our community has gone from believing in open source for controlling destiny, security, etc., from avoiding unnecessary corporate capture (think of all the rants about Swift!), from believing in heterogeneity and portability, to just saying, look, if you're not using the product of one of two loss-making AI unicorns to do all of your work, you're just being a contrarian weirdo who is doing it for performative reasons.
I guess if that is your definition of contrarian — that everyone else is doing it so why aren't we? — then I am.
But in my experience it's not so different to "why aren't you on social media?" — which was an argument from insecurity and the desperate need to fit in.
I personally believe in independent thinking. I have things I want to learn about how all this actually works, I have questions about my own life and future, and I'm not going to simply chuck them away because you or anyone else thinks it's just being contrarian and we shouldn't question the direction everyone is taking. So I am learning the tech with a stack I control, while continuing to do the job I do.
You can do what you want. You're probably right, and I'll be quite happily wrong and continue in my wrongness even though, based on how you characterise it, this kind of independent thinking appears to unsettle you.
How much effort did you spend to make the local setup work well?
Local models are not plug and play, you need to adjust settings to make them behave intelligently.
On my side one model looking super dumb right after install ended up being 'agentic' after using the right chat template, fixing how tool calls are setup, fine-tuning plenty of params values, and adjusting the whole setup to my own hardware. Claude.ai (free) was great at helping me sort that all out.
If that were really the case, basically nobody would spend so much money using them.
With my DGX Spark + 5080 + 3060 setup, I've tried all kinds of approaches, including quantization, acceleration, fine-tuning, post-training, and everything else.
It still can't compare to Opus 5.5 max in terms of speed, capability, and stability.
And when I calculate all the money I've invested, directly subscribing to the service would actually be cheaper, more efficient, and more stable.
Open source is great, but don't glorify it.
The architecture of the code I develop has never been simple. If you're someone who doesn't even work on things that need to be deployed to something like k8s, then basically, open-source models are indeed enough for you.
Found your issue.
There is a serious, malignant, dangerous disarray somewhere in our callibration of value, of worth, of what things are worth. Something, many things, are wrong here.-
But I trust the market to calibrate value about 10,000x more than I trust any person or group of people to do so. The thought of that is frankly terrifying.
Not that it's somehow "not people" (of course it is).
"Prediction" markets have rendered this clearly false of late.
> the reason I trust the market is the "skin in the game"
The market expects economic bailouts and other assistance during downturns, because individual investors with 401(k)s demand it, despite the perversions of market incentives that result.
Now people are disappointed with a thousand times that.
- Nervous about Redis being written in C? Switch to Garnet.
- Bun ported from Zig to Rust for $165,000 at API pricing -> https://bun.com/blog/bun-in-rust
- Microsoft ports Copilot runtime to Rust for $120,000 in API costs without blocking other development work -> https://github.blog/ai-and-ml/generative-ai/migrating-the-gi...
- Someone made a "clean room port" of Total Annihilation in golang, looks like it started in August and is now the most convenient way to play Total Annihilation. https://github.com/nanolathe-gg/nanolathe
- You only have to decompile / reverse engineer an existing game once and now you can have an agent occasionally add features.
Once you do the port and have worked out the regression issues, assuming you are switching to a safer language with better guarantees, you are now off of the old codebase and don't have to pay the costs of the old codebase. It's just a one time burst of token usage.
Costs of tokens will continue to fall as more hardware either becomes rentable (datacenter comes online) or moves into the used market, as workloads move onto users individual machine (because smaller models are becoming more capable and getting better inference engines and harnesses).
You can also task an agent with researching for "list the top 5 performance improvements we might be able to make" and have it research overnight, then the next day start a fleet of agents iterating through the findings that seem the most workable . Once you gain the performance win, that's basically permanent.
I think that token usage is a very elastic market, you can choose to spin up a fleet of agents or... you can stay human-in-the-loop, or you can type code yourself if you don't have a token budget.
So, yeah, I don't think Anthropic or OpenAI are worth much when hot on their heels are open-weight models hosted by major cloud providers + Nvidia and cloud providers themselves building their own models.
That said, Anthropic is an interesting company. I mostly use local models like Qwen 3.8 Next Flash, but when I occasionally use Opus 5.5 for difficult I enjoy the experience.
Some say that if Anthropic keep delivering improvements and be the first it will make it but for that, you need to disregard all competition from other llm companies and from other competitor systems.
And even though they have the smartest model ever, if someone offers something 90% of it for 90% cheaper it will be worth it.
We already see plenty of companies reducing ai cost because they don't see enough roi