h3lp

ARMbian is quite alive, and supports many RISC-V boards in addition to ARM64

Most of the SBCs are supported by fairly standard Linux distributions: Debian/Armbian or Fedora; you just boot from the approriate image on an SD card. Some SBCs have eMMC storage and/or a M2 connector, so you can either keep running off the SD card or transfer the image to that storage.

The value proposition of using SBCs is in their embedded connectivity; in addition to standard USB/network/HDMI ports they tend to have built-in connectors for:

* MIPI input, for video cameras

* MIPI output for LCD panels

* i2c and SPI for weird peripherals (accelerometers, temp sensors, etc)

* i2s for sound

* GPIO and timers/PWM for custom peripherals like motors/servos, contacts, programmable LEDs etc.

It occurred to me that tariffs are really a backdoor way of introducing consumption tax (aka sales tax, VAT, etc). It has been a conservative policy goal for many years, as the conservatives believe that income tax penalizes entrepreneurship, but the politics make it virtually impossible to switch to because consumption tax is regressive, and it is a huge change. The tariffs debate and the general political atmosphere created a misdirection: voila, we have both income and consumption tax! The next development will be the politicians discovering to their horror the high levels of taxation, and abolishing the income tax. Brilliant!

That's a great insight---the problem with LLMs is that they write code and elegant prose for us, so we have more time to do chores. I want it the other way around!!!

Folk medicine that succeeded in rigorous clinical trials is called 'medicine'?

On Banana Pi F3, which is similar to VF2, you can boot off SDcard, and transfer the OS to eMMC or M2.

Well the railroads aren't the seat of power that they used to be---the progress ebbs and flows in mysterious ways. Even their arguable replacement, air travel, is not the paradigm of economic dominance.

Everybody is now excited about the top end: more parameters, larger context window. Justly so, but I think that it's more important in the long term what will happen at the low end, both in the software and hardware direction. Software obviously has started (c.f. DeepSeek). As for hardware, Google TPU ASIC is six years old, and has presumably 4096 ALUs and uses few watts of power to perform 4 TOPs. Several years later, I would hope that it's reachable goal for those 4 watts to do as many TOPs as say H20 (400 or so).

Once the models stabilize, I could see a low power inference peripheral design with huge TPU matrix with an embedded model parameter Flash EPROM feeding it. This could go into anything from white goods to computers. Imagine a thermometer that you speak symptoms to and ask for diagnosis; is it a good idea? I don't know... but is it possible? I could see that it is.

It occurred to me that tariffs are really a backdoor way of introducing consumption tax (aka sales tax, VAT, etc). It has been a conservative policy goal for many years, as the conservatives believe that income tax penalizes entrepreneurship, but the politics make it virtually impossible to switch to because consumption tax is regressive, and it is a huge change. The tariffs debate and the general political atmosphere created a misdirection, resulting in both income and consumption tax being with us now. The next development will be the politicians discovering to their horror the high levels of taxation, and abolishing the income tax. Brilliant!

> There are a few regulations that have a high cost/benefit ratio (e.g. ban leaded gasoline),

I think you mean to say that it had a very low cost/benefit. Otherwise, citation seriously needed.

35M of 1937 dollars is $784M$ today according to https://www.usinflationcalculator.com/. Also, 97M$ is quoted as a total budget for maintenance; painting apparently costs just 0.3M$ (I have no idea what else goes into mainenance--probably they include roadways and connectors, toll booths(?) etc.)

Well, I don't really understand the detailed content of executables compiled by GCC/LLVM either, but I am not going to go back to writing assembly language. Having said that, I am old enough to remember worrying about compiler bugs, just like today I worry about LLM hallucinated vibe code. The hope is that we'll figure out how to make it more reliable---and I believe there seems to be a clear path forward.

Fracking was a brilliant invention, but may be reaching inherent limits---there are lawsuits between oil companies about fracking fluids from one well flooding and disabling other wells.

The AI discussion reminds me of the assembly vs high-level languages (HLL) debate of my youth. The claim was that expert assembly language programmers could write optimized code that was much faster than compiled code, and therefore HLLs will always be the 'second best'. Obviously, that was a misunderstanding: compilers got so much better that, on average, they beat human assembly programmers, especially on large code bases where careful machine language optimization does not scale. There's still a place for optimized assembly language programming, but it's exclusively in high-performance highly reusable libraries.

I think a similar pattern will emerge for AI: it'll be used for routine code, while freeing skilled programmers for algorithm development and similar high-level tasks.

A common FP4 layout is 1 sign bit, 3 exponent bits, 0 mantissa bits. There's just not that much difference in complexity between that and a 4-bit integer---the ALU can just be a simple table lookup, for both FP and integer.

thanks for the benchmark tip. FWIW, firefox 140.0.4 on Fedora 42 runs pretty much all tests at 60 fps or therebouts.

One large but forgotten effect of the dotcom bubble was an excess fiber capacity, that allowed smooth growth of internet in the following 25 years---average internet speed in the US is 200 Mbps, and a significant number of households is on a gigabit uplink. I take your point that GPU hardware amortizes away faster than fiber, but that's true of all computing hardware: the average lifecycle of a server is around five years.

I see an analogy to the discussions from my youth about compilers vs. assembly language programmers. It is still true that assembly is required to write high performance primitives, and that a competent assembly programmer will always beat a good compiler on a small function---but a compiler will consistently turn out decent and correct code for the entire project. So, basically, the compilers won, and assembly is relegated to be an important but niche skill.

Well, historically, we used to have a robust industrial R&D that even did basic research: Bell Labs (astrophysics, semiconductor physics), IBM, Xerox, RCA, Kodak and many other; even Ford had an R&D lab with physicists. Their record was quite impressive, as we all know. The weird trick that made that work was a tax policy that supported that. IIRC, R&D was expense as opposed to amortization, together with a laxer trust regime that allowed large profits that could pay for all that.