aprentic

It's always about logistics. The Three Kingdoms War was one of the bloodiest conflicts in human history. It was largely enabled by the invention of the wheelbarrow.

I like Cursor but that valuation is a hard sell.

Their valuation should be very closely tied to how how many tokens it takes to get from Void to Cursor.

If those values diverge by much, something is rotten in the state of Denmark.

You're essentially saying that privacy violations are baked into the cores of these systems.

It might not meet some pedantic definition of "Malware (tm)" but, for all intents and purposes, it fits perfectly.

Why not?

It's a software feature designed to benefit Google at the expense of the user.

The obvious answer to behavior like this is warnings that escalate up to a sitewide ban.

When a human is abuses a system, that human normally loses access to the system.

It seems that EVs and ICEs are sufficiently different that traditional car companies don't really enjoy nearly the encumbrancy advantage they once had.

The top EV manufacturer started as a battery company. The second place EV company started as an EV company.

Various ICE manufacturers have spent decades innovating and refining ICEs and building logistics chains optimized for ICEs.

The big problem that the high end ICE manufacturers have is that the things that made them special in the ICE market don't apply as well in the EV market.

You can potentially justify 6 figure price tags if you're in the luxury market. Hire famous designers, pay for premium materials, and leverage your brand name. If you want to sell cars for 6 figures based on performance, they actually need to perform significantly better. There are a bunch of much cheaper EVs that have better performance.

Just jacking up the price and relying on conspicuous consumption is how you get the Fyre Festival.

What if the framing were slightly restated?

I view all three of these more as good advice than demands. As I see it, people ignore this advice at their own peril.

It would certainly be nice if we could craft some system that removes the need for such advice but until then, I'll try to follow it.

I see this as the same dilemma that comes up around personal physical safety. In a just society, nobody should have to change their behavior to protect themselves from criminals. In the real world, people who ignore this advice, are more likely to get mugged.

The nice thing about statistics and math is that you don't need to stop at a feeling. If doubt their math, do it yourself.

I still have my G-Shock from the late 80's. It has a crappy light, a quiet alarm, and limited functionality.

But it's damn near bullet proof and I change the battery every few years.

I like this. It's an inversion of the old addage, "a poor craftsman blames his tools" and the corollary, "use the right tool for the job" (because a good craftsman chooses the appropriate tool).

You don't get to bang on a screw and blame the hammer.

Reading between the lines of the article it seems advanced but not too surprising.

I assume that at night when it "withdrew to a covered location" there was opportunity for maintenance, battery swaps, etc.

The article says that it successfully carried out "multiple calls for fire." That sounds like over those 45 days there were multiple missions to provide suppressive fire. They're not explicit about what that means but it sounds like, "if you see anything moving in this arc, take a few shots at them". Presumably there's some AI to prevent it from wasting ammo on really dumb decoys.

A "simple" mobile automated turret has been around for a while. The novelty they would be demonstrating is essentially battlefield robustness. They aren't claiming that this machine can operate completely autonomously for 6 weeks but the incremental pieces are still hard.

Do you know if that's true of non-English models?

As I said elsewhere, Deepseek injects Chinese characters into responses. Anecdotally, that seems to happen when the context gets longer. That suggests that they're primarily trained in Chinese and I would expect them to use fewer tokens for Chinese than English.

Deepseek will regularly spit out Chinese (汉字）during English sessions. They generally seem to be syntactically related but it makes me think that there's some overhead of using English with an engine that's primarily trained in Chinese.

There's certainly an interesting question here, even if Tokenstree doesn't provide a solution or even define the problem well.

The broader questions are still interesting.

If an AI is trained more on language A than language B but has some training in translating B to A, what is the overhead of that translation?

If the abilities are combined in the same model, how much lower is the overhead than doing it as separate operations?

ie is f(a) < f(b) < f(t(B,A) ? where a and b are in A and B and f() and t() are the costs of processing a prompt and the cost of translating a prompt.

Then there's the additional question of what happens with character based languages. It's not obvious how it would make sense to assign multiple tokens to a single character but there's the question of how much information in character based vs phonic based words and what the information content of sentences with either one is.

I love the idea, in principal, but I think it's impossible in practice.

A good strategist makes the outcomes of individual battles predictable. That makes it terrible for unit players.

I used to play Planetside 2 with a very organized group. Winning was fun at first but you were ultimately a cog in a well oiled machine so it got old fast. It probably got old even faster for the other players who were just trying to play a regular fps.

Those videos are awesome! 3B1Bs visualizations finally made e^(pi*i) make sense.

His videos on Euler's formula inspired me to make a silly toy so I could play with it myself.

https://gitlab.com/aprentic/complex-viz/

Is IBM really the company we want to hold up as a moral standard?

https://en.wikipedia.org/wiki/IBM_and_World_War_II

It's hard to tell without the spec sheets.

The magnets are almost certainly from China.

The top copper producer in the EU is Poland so that's a possible source of copper. They're pretty far down the list though so it's likely that a large part of the copper is coming from places like Chile (top producer in the world).

You could counter with that or you could read what's actually in this sub-thread.

"Some rich guy in the Netherlands" isn't about being a nice target. You keep saying it works for you but you can't demonstrate any way that works for others.

I can point you to a number of places that sell any number of Chinese drone parts that don't involve a "1-900 number". You can find them on Amazon. Any number of drone vendors sell them through normal sales portals. The manufacturers will ship them directly.

A handful of companies that require a bespoke procurement process and are operating at a tiny fraction of the scale do not have any appreciable impact on the market for drone parts today.

You claim it works for you.

Can you demonstrate a process that others can follow?

This whole thread is in response to an attempt to someone trying to source parts to manufacture something in the US.

If "for me" is limited to some rich guy in the Netherlands, that doesn't solve the problem "for anyone else."

Can you show me? Is this some privileged access that you get as an investor?

Its easy to verify that Motor-g does not ship outside of Ukraine. I just put 4 of their motors in a shopping cart and tried to check out. The drop down menu for destination country has a single option, Ukraine.

Arctus does not list a single price on their website. That's also easy to verify. Every single product on their website only says, "request product data", or "coming soon".

I can't install a motto in my drone. None of the alternatives will allow me to put a physical drone part in my hand with any degree of reliability.

I checked both the links in the other comment.

While they satisfy the technical requirement of, "there exists an alternative" neither of them is generally available as a viable alternative to China.

Motor-g doesn't seem to ship outside of Ukraine. That's totally understandable but for anyone outside of Ukraine, they effectively don't exist.

Arctus asks you to contact them just for product info. It seems they just raised 2.6M in seed funding 3 months ago. It's great that there are startups in NL but that's not even close to a replacement for China's scale yet.

Both of these may change the landscape in the future. For now, neither of them is a practical way to get drone parts without China.

Do you have any links?

How?

As near as I can tell, the vast majority of the parts are made in China. When I look at the few alternatives, they're full of Chinese circuitry. If I look at circuit components, they're all made of Chinese raw materials.

Both Ukraine and Russia are planning to deploy (and use up) several million drones over the next year. Iran just joined them as a major procurer.

Where are all the US and EU component factories?

The thread is about manufacturing in the US so tariffs do have to be factored in.

On those specific parts:

Motors: T-Motor F90 1300KV - $119.60(incl shipping) + Tariff

ESC: Holybro Tekko32 F4 50A - $88.97(incl shipping) + Tariff

FC: Matek H743-SLIM V4 - £88.12(incl shipping + VAT) + Tariff

Radio: Radiomaster M2 $95.99(incl shipping + sales tax)

The FC was from a UK store but it originated in China. I already had the radio so I don't have current prices on it.

I'd love to find a list of vendors that have comparable parts, in stock, and without being insane multiples of those prices.

edit: formatting

I suspect its easier to find Greenlanders willing to do that than it will be to find Americans willing to work in that mining settlement.

Unless we go full evilmode and just run them with slave labor.