I have a number of colleagues in Europe that normally I occasionally see. Since ICE ramped up last year, none of them have set foot on American soil at all. No one wants to risk even legally being in this mess.

And is also, according to Trump at least, also not the acting president. The acting president according to Trump is… Trump

No, she was the second in the regime.

The Trump admin felt it was probably more effective to give the current regime a slap to let them know who is calling the shots, but otherwise leave the regime largely intact. If they can just get the same Venezuela as it already existed, just as a vassal state, mission accomplished.

Versus Machado, who was the opposition leader trying to ingratiate herself with Trump up to giving him her peace prize in hopes of having the US implement her indirect electoral victory in 2024. Trump administration has shown zero interest in making that happen.

At least taking their figures at face value, about 75kwh to generate a gallon of gas, and let’s say 67 kwh to get a an EV 200 miles (assuming some losses between the generation and the actual battery capacity, and 3 miles/kwh which is on the low-ish end of EVs, but realistically close). The most aggressive hybrids getting 50 mpg so we end up with it being about 4x worse than charging an EV with that energy source.

At least at residential rates where I live, that’s about $10, so it would only really make sense when gas gets to $10/gallon, otherwise, go to the pump for the fossil fuel. That’s ignoring the cost of the station itself.

So maybe nearer than one might imagine, but still highly impractical. Maybe if they doubled the efficiency and gas eeks up without residential electricity rates going up…

But all this is assuming it will work exactly as well as they say it is, and I’ve learned to have a healthy dose of skepticism… Here though I can be as optimistic as they like and it’s still a tough sell…

Well, I don’t know if the reason given is that significant, they’d just plan around the fixed weight. The issue being the energy per unit volume/weight being so far behind hydrocarbons that some applications do demand it.

So while stationary/grid applications may lean battery since size/weight hardly matters, and EVs are debatable good enough for many scenarios, I will grant that for aircraft, boats, and some heavy equipment it’s hard to beat hydrocarbons.

Unfortunately, on that front it has to compete with extracted hydrocarbons and doesn’t seem like it can compete as yet. It however may give hope for a more resource constrained future that the battery-hostile scenarios may still be fulfilled in a sustainable way, just at higher relative expense than today. Or they iterate on their processes to have cheaper equipment and/or increased efficiency to come closer to competitive with extracted hydrocarbons. Or a viable thing to reference for some governments mandating sustainably sourced hydrocarbons when they are really needed.

How much do we have an over generation problem in general though? I suppose the argument would be that solar is curtailed because they don’t want to deal with the potential for overgeneration, but we already have a number of approaches for energy storage. Their pricing for generating at most a gallon a day is a price exceeding a battery system of LFP that could do a lot more than a gallon of gas. This is ignoring the rather significant potential of Sodium batteries.

So this doesn’t look to be cheaper than battery systems, it looks to be way less efficient than battery systems. The biggest use case as energy storage in general seems to be if you want it to spend a few months (but not too many months, fuel degrades in the tank after all). The more narrow use case is to cater to scenarios where you absolutely need the energy density of gasoline, so boats and airplanes critically so, maybe some heavy equipment. I’ll grant that, but if particularly sodium batteries will be an acceptable approach, it’ll be better than this solution in that very wide variety of circumstances.

Even then, the value prop is questionable.

It treats sustainable energy dedicated to this purpose as “free”, ignoring the opportunity cost of using that energy directly.

For example, let’s say I dedicated my solar exclusively to making gasoline. I could get about 14 gallons a month of “free” gasoline… Except my home power bill would go up about 150 dollars a month… opportunity cost would be over 10 dollars a gallon…

Problem is that the efficiency is on the ground here.

The same energy that might get an EV 200 miles instead produces a single gallon of gasoline, to get a sense for the relative value of the efficiency.

Note that this demo takes things over 100ghz. So the challenges associated with mmwave (and wigig) are even more.

They hint at their goals by mentioning fiber in a datacenter, where they are now getting to 400/800gb speeds, so in the ballpark of this demo, but this would be a shared medium instead of a switched network, so it’s DOA there as well.

Unfortunately, looks like the breakthrough is silicon that can credibly work with those frequencies at all with a reasonable power budget, by simplifying and reducing power draw. Maybe it could somehow reduce energy usage of wifi, but they seemed to be all in on being over 100ghz… So the tech won’t be increasing the throughput of anything more practical.

It pretty much would be line of sight only.

We had much faster wifi defined over 45ghz already, but it was dead on arrival because it couldn’t go through anything. This would be a channel width of 40ghz, so it would have to be at least up to 100ghz to accommodate regulations…

Well as long as you never turn around and put your body between the headset and it’s wireless peer.

Note that 802.11ay to get 20-40gb (approx 2GB to 3GB/s) is a thing, and it’s ignored because going over 45 ghz is just impractical. This experiment would have to go even higher than that.

ROM because ultimately they have to support the device

They also make deals with companies as to what their preload has, what the preload forbids from being uninstalled, and what subscription services for into their platform.

While optional, it is also the default behavior.

It’s a bit directly on Microsoft, unless you go out of your way, bitlocker will upload the keys to Microsoft. They assume you want them to help recover your data if your tpm becomes unavailable.

Interesting fun fact, when I tried to swype type bitlocker it really wanted to put bootlicker instead.

My guess is that they have email hooks into LLM and call each entry point into LLM invocation an ‘agent’ and I have seen in a lot of companies the easiest way for them to have an email is to just add them to the directory.

It’s still dumb as hell, but I am no stranger to non-human’s in an ‘employee directory’, though usually it is supremely obvious that it is a non-person so if it’s at all confusing it means they are being ‘cute’ about their accounts.

Yeah, very good analogy actually…

I remember back in the day people putting stuff like ‘Microsoft Word’ under ‘skills’. Instead of thinking ‘oh good, they will be able to use Word competently’, the impression was ‘my god, they think Word is a skill worth bragging about, I’m inclined to believe they have no useful skills’.

‘Excel skills’ on a resume is just so vague, people put it down when they just figured out they can click and put things into a table, some people will be able to quickly roll some complicated formula, which is at least more of a skill (I’d rather program a normal way than try to wrangle some of the abominations I’ve seen in excel sheets).

Using an LLM is not a skill with a significant acquisition cost. To the extent that it does or does not work, it doesn’t really need learning. If anything people who overthink the ‘skill’ of writing a prompt just end up with stupid superstitions that don’t work, and when they first find out that it doesn’t work, they just grow new prompt superstitions to add to it to ‘fix’ the problem.

Unless Nvidia gets in on remanufacturing, those GPUs are never going to repurposed for residential usefulness. B300 was designed from the onset for datacenter ai use exclusively, with no concessions for theoretical video out integrated to a board that overall demands over 15 kw. You couldn’t even power it with a 60A 220V circuit.

Some of the storage could get more consumer support, SAS is unusual but if there were a glut then various solutions could emerge. Similarly EDSFF cages aren’t really a hot consumer item, especially not e1.l, but I could imagine a glut driving home friendly adaptions.

DRAM modules are somewhere in between, though practically speaking they won’t be workable outside of their initial application.

There was a time when home and datacenter got closer together, but there’s been quite the divergence the last few years.

Yes, because Microsoft’s revenue growth is in fact the most important thing the folks at Davos had to think about…