Every time people hear what it actually involves is outrage and opposition, at least all the way back to when Intel first announced that they wanted to hardcode identity numbers inside CPUs in 1999. But the corporations involved have been relentless, and have sunk countless billions of dollars steadily forcing it forwards, and building front organizations to obfuscate and whitewash it.

With Windows 11, Microsoft was finally able to FORCE Trusted computing hardware into every new computer. Not just every new Windows computer, but every computer.

Every new Intel PC CPU has built in Trusted Computing enforcement hardware. Every new AMD PC CPU has built in Trusted Computing enforcement hardware. Every new ARM PC CPU has built in Trusted Computing enforcement hardware. The only processor lines that DON'T have it are the microcontrollers.

There are some unlocked smartphones available, but as far as I can determine it's literally impossible to buy a smartphone that doesn't have hardware trusted computing built in.

I specified that enforcing Trusted Computing at the internet access level is still a long term goal, they couldn't get away with it today. However they are well on the way to success. Virtually all new hardware supporting trusted computing, all the front groups rolling out standards and systems, and as it gets incorporated into things everyone is going to increasingly running into situations where they get locked out of stuff if they're not Trusted Computing compliant. Streaming already restricts you to the worst quality if you're not compliant, and it's only going to get worse as pre-Win11 computers fade out and as more things require it.

Yep -- but there's no systematic routine national program, that's the issue.

There's a bunch of quite specific UK-market problems. Australia also has marginal pricing, but it has 5 minute dynamic segments, and the market is set up in such a way that it consumers can see the effects in real time and thus modify their behaviour to respond to the price signals. Here in Blighty, we just get a bill months after the fact, about which we can do nothing.

What saddens me was that in the course of a 30 minute chat with Claude, I was able as a mildly informed layperson to work up a suite of pretty plausible reforms that would deliver a much more functional market. Both politically and technically plausible. I think so much of this is about will to act, and maddening British negativity. All that sucking of teeth.

I'm Gen X. I'm white. Both are important to my point.

So when dealing with white people over the age of 70, the vast majority are huge Trump supporters and watch Fox News all day to see how the Democrats are "trying to destroy America". All of these older supporters think he's been great and truly are incapable of understanding why anybody rational doesn't see that Trump has been great.

With my generation, the schools have always been integrated. I have always lived in integrated neighborhoods. Maybe I went to majority white schools and lived in majority white neighborhoods, but neither prevented black people from going to those schools or living in those neighborhoods. Gen X is really the first generation of truly integrated America. I went to a pretty large high school and I would estimate that at a minimum 80% of my living white former classmates are huge Trump supporters who vehemently hate all Democrats. Yes, plenty of people who voted for Trump remain 100% committed to him.

A key part of the MAGA movement is absolute shamelessness: this is true of the leaders and the members. They are shameless about their own lies, about holding themselves to different standards from others, about being fascist, about being racist, about hurting other people, about hurting themselves, and above all, about being stupid. They don't merely take pride in being shameless; they recognise that being shameless opens up a wider set of actions to them at any point in time than those that would be available if they allowed shame to limit them. And they recognise that their opponents will spend time pointing out their shamelessness, expecting that to have an impact, rather than accepting it as an amoral and pragmatic decision and fighting them as they are, tooth and nail. So it's been very useful to them. Of course, there are risks to being shameless: your actions may lose you popular support, for example. But that is why part of their program of shamelessness includes seizing control over speech and education, and above all, of cheating in elections and courts. They want the question of whether they have popular support to become irrelevant.

I'd hardly call exact date of birth "low information content".

And yes this does reveal exact date of birth, regardless of the bullshit obfuscation that it supposedly only reports age range. The server simply tracks the reported result every time the user connects, and on some specific day the result CHANGES to announce their date of birth.

-

This is based on SLSA (Supply-chain Levels for Software Artifacts), brought to you by the same fuckers making Trusted Computing and the TMP (Trusted Platform Module). It's part of the same shitstack to prohibit you from altering your software and to lock you out of your own files, and to send spy reports out over the internet so you can be cut off if you "fail" the Trusted Computing check.

The software can use a TPM's (Trusted Platform Module) Sealing function to encrypt your data such that it's impossible to access your own data if the software is modified. It can then pass control over that data only to software updates that carry a signed security certificate from Broadcom (or any other company using this system).

It is no longer open source, your system no longer works as you can no longer access your Sealed data if you change so much as a single letter of the code. Even recompile unaltered code won't work, unless you magically manage to get your build environment absolutely identical to the company's build environment and get byte-for-byte output. Even that may be impossible with the newer levels of non-deterministic compiler optimizations.

Also, with TMP's Remote Attestation feature can be used to transmit your machine's software configuration over the internet, so that you can be cut off if your system doesn't match Broadcom (or other company's) cryptographically signed certificate.

And then of course there's Network Access Control (NAC) / Trusted Network Connect (TNC). In the long term, the goal is for ISPs to use NAC/TNC to interrogate your computer for Trusted Computing compliance, and deny you any internet access whatsoever if your machine isn't compliant. Software with this sort of "security" certificate would pass inspection, while any attempt to alter the code would be detected as "tampering". You then get "quarantined". What "quarantine" means is that you are denied internet access - with the exception that you do get very restricted access which can only be used to download the approved software to "fix" your computer into Trusted Computing compliance.

I think this is likely correct. Earlier in the year my investment club had its monthly meeting and at the time Trump was making another threat about Greenland. I told them that I expected that every year that Trumps remains in office, I expect him to make a threat about taking/invading Greenland, probably in January. At the time, I think there was a lot of talk about the Epstein files. I don't expect the US to actually invade Greenland, but the talk of doing so is hugely distracting to Americans and it makes them stop looking elsewhere.

It's an inherent risk. It can be engineered out to a degree, but it requires a program of PPM to actively manage the risk, because seals can corrode, fittings can come loose, etc.

Inference is neural network limited. On systems with entirely separated data and compute, this is a memory bus constraint. In quantized systems, ALUs and FPUs are illogical. A simple 16 cell LUT is suitable for noise free multiplication. A neural network circuit with a dedicated LUT and addressable buffers more or less eliminates bandwidth constraints. But this requires neural networks more similar to FPGA cells rather than classic compute.

That said, separating inference from data drops network depth considerably. And in transformers depth increases memory access exponentially which is why nearly identical MoE and dense models perform so differently. We of course need data centers for models with massive numbers of active parameters. But massive numbers of active parameters weaken models as that means depending on training rather than looking up facts. Huge active parameter sets will never be a good design. A real world analogy is that huge models are like Jeopardy champions who have crap loads of partial trivia facts in their heads. Smaller models are like librarians who don't have a trillions useless facts memorized but knows how to find the data for the researchers who will use the librarian repetitively to follow through to the answer. The model only needs enough training to use the data.

What really matters is, how fast it can find the data.

Inference suffers greatly when you treat weights as data. And bigger models have more trivia like facts memorized from their training. This means more active parameters and greater network depths with higher memory bandwidth needs. Even now, you see the chatbots are improving drastically because they rely substantially more on data and way less on inference.

I am not sure which part you consider gibberish, but it makes perfect sense to me. But it might sound better in my head.

P.S. I actually have considerable data to backup many of my claims. At work I'm sitting on about 50MW of compute. We build in shipping crates in a repurposed mine. There is a group of us where our goal is to avoid spinning up more compute. We need to deliver inference to 200,000 users eventually. (not customers, employees) and we could throw millions at NVidia and end up hosting some crap model, but we focus instead on cutting that memory bandwidth need. And yes, I'm far from being the smart guy on the team.

From here in the UK, this is just so striking. Here, we have a single design of number plates (more or less), optimised for readability, with a (more or less) uniform requirement to be on both the front and back of the vehicle. They are so legible that ANPR systems (=US ALPRs) don’t need clevercreepy AI to decipher them, and so they don’t. So we don’t have Flock, and the system is under the control of a public service with a fair amount of oversight, and while it is capable of being abused and has indeed been abused, the scope for abuse is much more limited.

See also: breathalyser rules; how police officers behave on traffic stops and on car chases (largely); etc.

I agree that there are actual ICE-specific failure modes that can cause a fire in a car that’s parked and switched off. It’s also true that the article got itself muddled by not talking about any of those!

There have been several recalls for this kind of failure mode:
* Ford Bronco Sport / Escape (2024): Ford recalled about 42,650 2022–23 vehicles because a fuel injector could crack and leak fuel into the engine compartment. NHTSA explicitly says the fuel could reach hot engine or exhaust components, creating a fire risk.
* Jaguar F-PACE (2024): Certain 2024 F-PACEs had an incorrectly specified plastic oil-filter cap that could crack. Pressurised engine oil could leak onto hot components such as the turbocharger and exhaust manifold, potentially starting an engine-compartment fire. JLR says it had received reports of a small number of actual fires.
* Kia vehicles (2025): A piston-ring problem could eventually damage the engine, including creating a hole through which engine oil could leak onto a hot exhaust component. Kia explicitly identified fire as a consequence.
* Volkswagen vehicles (2025): Another recall involved cylinder-head-cover screws that could loosen, allowing engine oil to leak. VW says that if the oil reaches hot engine components, a fire could occur.

I think it’s fairer to say that OEMs *can* use smaller “starter” batteries for EVs than for ICE, and that some, but not all, do. Tesla does. But Polestar, for example, still uses a 70aH in the Polestar 2. Prob decided it was easier not to re-engineer that system, although it’s obviously a waste

This seems smart.

The federal government needs
  a) more tax money to reduce their dependence on more bonds
  b) more tax money to redistribute to startups
  c) reshoring to avoid just being too far behind.

So, taxing imports is a great idea. It doesn't matter if the trickle down effect works or not. The government has bills to pay. The government is owned by the people. Therefore, the people need to pay their bills. These taxes are good... we need to reward Trump for finally bringing new socialism to America. And what is best is that this really will steal from the rich and give to the poor.

BTW... Little secret, if TSMC spun down there business right now over the next 12 months, we'd be fine. We might have to rewind our tech a year or two and it might take some time to grow capacity, but it really wouldn't matter. TSMC just gives us a 12-18 month boost over their competitors. If they scale down as Intel, GF, Samsung, SMIC and a few others scale up... it really wouldn't matter.

The absolute best data centers today should be written into history as a dark dark stain on the progression of humanity. It is a sign that no matter how stupid we have been up until now, we managed to find that hole has no bottom.

Consider this... The reason we need data centers that massive today is because the tech isn't ready.

NVidia B300 filled racks using 100kw each are barely capable of running modern frontier models. Even if we designed more powerful chips that could do the job, there isn't enough silicon wafers produced in the world to do the job. Even the most amazing semiconductor technologies on earth are not good enough. Consider that even if we could build 1TB HBM4 NPUs, the time it would take to generate a token on a transformer that large would be slow.

What we know

Transformers will leave the data center between 2030-2032 as it will never make sense to run transformer inference in racks. It's at least 1000x less green than local inference.

Intel, AMD, NVidia, Apple and every RISC-V/ARM chip supplier on earth is 100% focused on making local inference the thing.

We will soon stop training massive models that start becoming obsolete the moment the last parameter is fed. Even now, Qwen 3.8 27b is good enough for a year at least (first model to do this).

For LLMs, we'll just stop wasting cycles on huge networks and will focus more on stronger harnesses attached to better data sources.

Cloud AI will be more about acting as data providers. Models will become reasoning engines, the "smarts" will come from things like vector databases.

Vector databases will use kilowatts, not megawatts. Data storage will be much smaller.

Things like Intel's dead Optane product will become the future of datacenters. Google Search, Microsoft Bing, maybe Baidu are the best positioned companies for what comes next and they don't need 100KW per rack to deliver. If I were to enter this segment, I would pair a customized KLV object query engine onto a 1024 bit wide ecc memory bus and connect it with low cost networking (think Intel's Ethernet fabric tech). Then add a super capacitor to flush to flash on power loss... Or just hope Optane comes back. Superfast, massive, token based vector databases are the future of cloud LLM

What's next. In 3-5 years, every single rack in every single AI data center is trash

This is fact.

Even now H100 racks should be heading to scrap.

The chips are slow, they use old heat spreader tech, they get almost no performance per watt compared to B300. They are much harder to cool. They have so little VRAM it takes piles of them to run newer models. PCIe data center GPUs were always irresponsible purchases. SXM GPUs only run on special systems Noone but high end data centers can even run.

Consider that every single B300 or older equipped rack on earth is landfill in 3-5 years.

By then, the world will have overprovisioned RAM, we'll have moved 2-3 generations of cooling forward, we'll have designed denser racks, started employing all the awesome tech China built to work around lack of EUV... No really, Korean companies are already licensing CXMT tech for wafer stacking. We will have moved to lower fresh water depen....

You know what? It is going to just be cheaper to bankrupt the shell corporations who own the current data centers than to upgrade them to new tech. Besides, the environmental issues will be such a mess. Imagine trying to dispose of that much fr4 epoxy without causing environmental disasters? It's probably at least as much epoxy as the total mass of the twin towers.

I can keep ranting but I promise this, there isn't a single piece of tech in the most modern AI data center on earth that is worth using in 5 years... Not even the racks. It's all garbage.