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
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
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.
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.