svnt

Micron is the only one to bail, and they are the smallest footprint. No one else has bailed.

Samsung recently shifted some capacity from HBM back to DDR5 modules, freeing up around 80,000 wafers/month, so DDR5 margins are currently good enough to compete with HBM for wafers, or maybe they are anticipating a turn, or both.

If you are training on data labeled by frontier models, how do you expect to exceed the performance of frontier models, other than in the cost dimension by recognizing simpler problems and routing to cheaper models?

They do not need to be lies, but they may be. They may also just be a mismatch between the level a statement is generated at and the level it is consumed at, the “level” here being something we could call a system’s functional context.

Additionally, a statement is generated by a system and reflects itself a separate purpose in its generation. A system can exist without statement of purpose, so why was it made? The statement has an intent and an audience. It is therefore necessarily more constrained than the system itself, and for the systems we are referring to, it is markedly less complex than the system itself. It cannot hope to capture the full function, which is what can serve (or not) a purpose.

If you log and view your "Usage" page there is a blue banner near the top in a blue box saying:

> We plan to raise the overall pricing for DeepSeek API services in the near future, with a significant increase expected. Please plan your usage accordingly. The specific pricing plan will be subject to official notice.

There are few conspiracies where the vectors between "capitalist organization makes more money" and "user can't reliably distinguish tiers of product quality" overlap.

I replied to the user above that referenced marginlab, but I believe marginlab uses the API. It is possible (arguably likely, in MBA-land) that the API and subscription accounts hit different sub-models.

Even if they use a subscription account, surely Anthropic can tell which one it is.

Is this using the api or using a subscription, though? The incentives are different for each, and it isn't the least bit unexpected that they would maintain API access quality while 'optimizing' the subscription experience to improve their margins (or losses)

It seems to do really this you would need to crowdsource it -- users individually give the lab access to a body of subscriptions normally used by average people, and the lab occasionally runs some masked version of the task through on diverse accounts.

It's because really what they are doing is using chemistry to split a cell. Adamala is a chemist.

They have just bolted an unsynchronized physicochemical process onto the boundary of the cell. It doesn't coordinate with anything to do with the cell. Both cells don't get half of the dna. They built stochastic chemical scissors that only work if you make the cell less cell-like.

18,600 engineering years sounds impressive to someone because it is the bulk of a career for 1,000 engineers. But it is less than two years for 10,000 engineers. The depth of understanding really hinges on which version is closer to reality.

Meta Horizons World probably puts up similar numbers if you sum up the hardware/software tech stack to get this number.

It has very human aspects, such as the beginning. So people switch off. And then it has long stretches where it is bulked up by claude and opus-4.8’s obsession with “honestly true,” “narrower” claims, how concepts “rhyme” etc.

I guess it is also possible this person has internalized claude, but I think their writing pattern is: short pieces: fully human voice; long pieces: ai-supported.

As to my personal views, I am sad to have lost the emdash and the antithesis, among other things, to the llm-cliche dustbin.

Right, this is cyborg text. The opening is in human voice. I am acknowledging that. Then it ebbs and flows over llm rocks.

The piece would be maybe 6-8 grafs without claude, and much of claude is papered over.

The one they claim is in their voice [1] has zero of the tells I am referring to. You can compare and contrast it with your fully/generated reference 0 (although that is fable and I have little experience reading its writing it seems to use many similar comventions) and the title piece and see what I mean.

[1] https://wilsoniumite.com/2026/06/07/labor-is-a-market-distor...

Parts of it seem not to be, but the bulk of it is. Here is a particularly clear example of opus-4.8-speak.

> Now, I want to be careful here, because this is the part where it would be very easy to start waving my arms around. Every single one of these has many causes. Suburbs and cars. Television, and then phones. A long list of things that have nothing to do with me at all. I am not going to claim I have found the one secret root of loneliness, and you should be suspicious of anyone who does. We cannot cleanly untangle these. That is just honestly true.

It is mostly written by llm. “narrower” and “I want to put a fence here” hedging, etc. This is very 4.8. Maybe llm that has been somewhat massaged by a human to sound less ai.

Founders are often synonymous with their businesses. In the case of celebrities the relevant exploitation is carried out by others in the value chain, they are just the lure/source.

It makes the government stupider so there are more excuses to bring in better private solutions.

Handicap the public services if they are working well, then talk about how bad they are to justify for-profit replacement.

Or don’t and just exploit the gaps directly with better private data, whatever increases proximate wealth inequality.

They are all about science and research. What they don’t want is for scientific discoveries to be publicly available, because then it is harder to leverage them for absurd profit margins.

Which they anticipated, which is why they were flagging and dropping back to opus on anything they could even potentially be called on.

Not sure but a green screen version of AWS for $500 is totally plausible and a decent metaphor for what you’d get with AI anyway.

I have found that on long-running tasks, many of the communication (and other) invariants get dropped at seemingly arbitrary points along the way. It probably just stopped doing the log.

As I see the perspective of a typical vp champion/customer/middle manager at a large org that buys their services: if my vendor increases their prices that is usually an indication they have something of value and things are going well. It is a negotiation, buy or build business choices, not embarrassing. If my vendor doesn’t raise, and then they cease to exist, or become so weak that a competitor of mine can buy them out, then I am actually in an embarrassing place.

My main job in this conception is to not be in an embarrassing place, and grow my budgets and headcount. Prices rising can even help with this, perversely.

They are because they can, because it serves as social proof, which convinces their customers that they are doing something of deeper value. Then in reality they will use it to develop channels preparing to use their customers (and the data customers trust them with) as the product in the future.

The play is the same as it always was, I assume: your data is the long term product.

IIUC in this particular instance of corporate <-> government space warfare, orbital decay should clear out the debris relatively quickly.

I agree with that as a premise, but again it seems to me you are selectively jumping way into the end game. There were early networks that did not standardize, and these nonstandard networks had advantages, and some of those advantages were sacrificed in market-driven standardization.

Intelligence must have interfaces, and those can be standardized. Businesses will try to remain provider agnostic, which will also drive standardization via standard sales and marketing methods.

Separately, we are doing our best to standardize performances on benchmarks.

I don’t disagree that right now transport of standardized mobile data vs emulation of human intelligence is qualitatively different, but perhaps primarily because it is early in development, and our vantage point this time is relatively from within the network, instead of outside it.

It isn’t animated at all for me?

I agree with much of what you’ve written but think you are missing the correct alignment of the mobile data timeline — mobile data had standards because it was forced to. It was forced to early because it was not a fundamental innovation, telecom itself was the fundamental innovation, mobile was a constraint relaxation. Intelligence might be forced to have standards as well, we will see what form the regulations take when prices reflect costs and healthy margins and become existential threats for many businesses.

This argument sounds nice at a surface level, but seems deeply incoherent and wrong in a way I haven’t yet put my finger on precisely. How can you say this:

> We have souls. We have an inner self. We can treat ourselves as transcendental subjects in the Kantian sense. We have emotions and spiritual life; we experience qualia. These entities are not obtained by addition to a physical state, but by subtraction from a complete physical account. Mental processes are physical processes described in a way that captures only their salient characteristics.

followed by this:

> The reason why this picture is more credible than any dualism is not that “science explains everything” — it doesn’t — or because “physics explains everything” — it does so even less. It is because of the hundreds of years of astonishing and unexpected success of the sciences that have convincingly shown that apparent metaphysical gaps are never such.

followed by this:

> Earth is not metaphysically different from the heavens, living beings are not metaphysically different from inanimate matter, humans are not metaphysically different from other animals. The soul is not metaphysically different from the body. We are all parts of nature, like anything else in this sweet world.

So it isn’t describable by physics but it is only physics? And there are no closures or gaps? Ok sure in one sense we can say everything is connected, but this article seems to me to demonstrate effectively that without these divisions, pursuing understanding of it is essentially intractable.

He seems to be describing the dissolution of some construct in his worldview that I am having a difficult time relating to. Anyone have a different take?

Closest thing is probably Bernardo Kastrup’s analytic idealism.

This is essentially where the field is going, started with predictive processing by Friston with his free energy principle, combined into Hofstadter’s “I am a strange loop” and then the continuing thrust with applying the FEP as well as Rosenthal’s higher order theories and Graziano’s attention schema theory.