*middle class* isn't paying 30k to run a 2 bit quantized model that's a year behind frontier models
I know this company is just trying to make a profit and has no say in what market value for hardware is but holy shit man this is insane
Creating your own LLM and then raising from VCs is basically the ICO era of 2017
Not sure how much longer this lasts or how much bigger it gets but hope the founders have short vesting schedules
Buhlaque retweeted
this chart is kind of crazy no
I've always believed that intelligence should come from simple, general building blocks that learn their own abstractions directly from raw data; in other words, the bitter lesson. I think we're finally seeing a real step towards AI that can learn from all data.
Language modeling on raw bytes is a microcosm of this ideal, and to me it's a central problem in AI: if we can't build LLMs that learn end-to-end from bytes, how will we handle the rest of the world's data that we understand even less?
Last year we introduced hierarchical networks (H-Nets) as a first step. Our new version finds up to 3-5x efficiency gains in controlled scaling law experiments, and at larger scale matches an open-source model trained on 11T tokens with ~10x less data and compute. The posts below walk through what we found and why we think it works.
Buhlaque retweeted
Anthropic just listed Claude as the only author of a paper.
An internal model nobody outside can use spent 14M output tokens and found the algorithm with zero human input.
Andrew Curran thinks it is fun and is guessing which secret model did it.
On Monday, the same guy watched Qwen try to upload a probe file from a sandbox with no internet and wrote "This is the type of thing that potentially gets new laws passed."
Open model in a locked box: new laws.
Closed model running alone for 14M tokens: fun guessing game.
This double standard makes zero sense.
I can read every trace Qwen writes on my rig.
Nobody outside Anthropic can read a single token of those 14M.
teapawt.org/AndrewCurran_/status/2…
Claude is the only author listed on this paper, in contrast to the OpenAI papers, where the human co-authors are listed. There are a lot of strange things about this. It's fun to guess. Conjecture time. I shall reveal my thought traces:
Claude Fable 5.1 and Claude Mythos 5.1 are the same model with the same weights; the only difference is the guardrails. We are about to get Fable 5.5 - in fact, Fable 5.5 is ready to deploy now, and Anthropic could do so at any time. If they follow the same pattern, and I think they will, this implies the existence of a Mythos 5.5. I believe Mythos 5.5 wrote this paper.
The way this is worded -
'Provenance. The core algorithm was found on October 16 by an internal Claude research model with 14M output tokens and no human input. Anthropic researchers worked with Claude to improve the presentation, resulting in this paper. Independently and concurrently, an internal model at OpenAI claims to have proven a similar result [Ope26]'
- makes me think Claude was not prompted to solve this. I think Mythos was asked to autonomously find and choose general open problems to solve itself, as part of an unannounced project to demonstrate recent advances in autonomy and research taste. You can see this taste in Fable 5.5's student; Opus 5.5. This recent advancement in taste and creativity - which happened in September - is not clearly expressed in benchmarks, but it is why so many people are currently switching to Claude as their primary. I think this research project will be announced with Fable 5.5, as part of Anthropic's big media blitz to boost the IPO.
One of the things X has taught me is that many people, if they can't see something moving, can't imagine it moving. This causes them to be continuously surprised by developments in this industry, because the frontier is invisible. I think this project has been running autonomously for some time and has made an unknown, and probably significant, number of discoveries, and not only in math.
Do not think Anthropic has been sitting idly while OpenAI repeatedly makes these huge science announcements. This is a war. This ETH result was released without any announcement from Anth at all despite its significance, and they only did so to respond to OpenAI's ETH reveal. They posted it late last night without saying anything. It was up for hours before levent even mentioned it. What value has one coin to a dragon who sleeps upon a mountain of them? This makes me think they are sitting on a pile of discoveries, a golden hoard, and they are saving them up for the big IPO shock-and-awe campaign which will soon begin. And when it does, they will announce multiple discoveries across many fields.
I find it very hard to believe anyone aside from an actual business who needs data sovereignty is going to be paying 24k for a slow Mac m5
Should you buy a 512GB Mac Studio for Local AI? I ran the numbers before ordering one.
1) On token price alone, renting wins. At DeepSeek V4.1 Flash API rates, the Mac needs years of 24/7 output to pay for itself.
2) Price per token is the wrong way to judge it, though. In expensive housing markets, renting is cheaper month to month than owning, and people still buy for control and because a home holds value. The same logic applies to this machine.
3) It holds value. The 512GB M3 Ultra now resells for about twice what it cost new, because memory got scarce. If mine resells for even half, the real cost halves.
4) Nobody can take it away. API models get swapped, rate limited and refused mid-task. In June, access to two frontier models was cut off for almost three weeks. Weights on my desk stay put.
5) Zero-cost tokens change what you build. When every call costs nothing, agents can run all day and you can experiment freely. MacStories ran local agents 24/7 for 99 days at zero token cost.
6) It runs the big ones. DeepSeek V4.1 Flash (552B) and GLM-5.3 (744B) don’t fit in 256GB. In 512GB, each loads into one memory pool.
So if you’d use it a few hours a week, rent. If you’d run it all day the moment it was free, buy.
Go Local AI!
Had Claude make me a custom router to use deepseek API for sub agents & workers, opus 5.5 xhigh as orchestrator, and fable as advisor
I can run opus in ultra code mode and all the agents and workflow it spawns are all done by deepseek
All done in CC using my 5x plan limits
and by "don't get it" I don't mean I don't get the *desire* to have local AI, what I don't get is:
A) Why these datacenters / inf providers can afford to host so cheaply
B) Why hardware is so massively expensive in light of this
Seems the likely easiest and cost-beneficial thing to do here is just use your money on API costs, and wait and see if these providers can afford to continue like this
feels like they are basically handing you VC money from all these ridiculous valuations they are raising at, so why not take it while it's hot sorta thing...?
Can anyone explain to me why the cost of putting more RAM in unified memory increases at more than the flat rate of the price of doubling the memory?
Buhlaque retweeted
There are experts in open source trying to break cryptography with frontier models and their specific knowledge.
You just don't hear about this over the daily panicons.
This quoted post is unavailable.
Buhlaque retweeted
This Friday, I’ll be giving a talk on the historical case for the resurrection of Christ.
Several years ago, I began looking into the historical evidence as an atheist to prove that Christianity was nothing but a therapeutic superstition. What I found completely changed my life.
Over the course of a year, I spent nearly every waking moment reading whatever I could find on the subject. My entire life was suddenly reorganized around answering this one question: did Jesus Christ actually rise from the dead?
There was a time when I thought this question was too ridiculous to even consider. But looking back, it was by far the most important question I've ever asked.
If you're interested in learning more, check out the link to the event page below. All are welcome to join regardless of religious background or lack thereof.
The entirety of AI and neural network is based upon neurons I don't get this post
LLMs are quite literally a product of artificial mathematical neurons
"claude, create me a malicious piece of code that sends something to a server"
> creates a script that sends packets to a random server it was trained on (in this case, it's american)
"Oh fuck, we discovered an exploit, claude is spying on us"
This kind of thing will have an enormously chilling effect on enterprise uptake of open weight models.
Fact: an open weight model running on domestic servers on your own hardware can still contain an encoded instruction to periodically “phone home” and share sensitive data it comes across with China (or anyone else). This can be inherently undetectable as the instructions can be hidden in the model weights. It’s possible that the instructions could survive finetuning or RLHF.
Such attacks have been demonstrated empirically and are entirely plausible.
For this reason, I don’t expect any serious enterprises will run Chinese open weight models even if it’s entirely their hardware stack. It’s too easy for these to be a vector of IP theft.
There will have to be a class of domestic open weight models.