hasmanean

I remember being excited by the potential of apps when I heard about Symbian…but some part of me knew that any product named Symbian would never succeed. It’s the type of name you come up with in round 3 of a name-brainstorming session, when the perfect name will only be found in round 10. They obviously gave up too early.

They brought a plastic case to a machined metal fight.

HW is hw and sw is SW…and never the twain shall meet.

Hardware problems are low dimensional ones…”optimize this variable (usually clock speed), make this block functionally correct.”

Software problems are high dimensional ones, especially the problem of designing an API the community will adopt. “What set of functions can I provide that will do X…and a subset of X…and a potential superset of X, on hardware generation H and H+1 and so on…all while taking security and OS privilege levels into account…” Oh and we must bow to aesthetics and programmer fashion as well as getting stuff into their hands as fast as possible…plus tools and documentation…plus outcompete everybody else in the world trying to steal the same devs as us.

Sw is what it is because it has so many degrees of freedom.

So the breakthrough happened when Apple shrunk the computer into a phone, not when Nokia grew the phone…? Interesting.

I remember writing apps for the iPhone and the biggest hurdle to getting it submitted was sizing the screens for the display variants.

There would have been no chance of developing an app for the Nokia ecosystem with all the different screen sizes your app would have to support.

Even downloading an app or ringtone required going through all these scammy websites and you had to enter your exact Exact model number and pay somehow…then they would text you a link or something.

Apple really cleaned up the ecosystem with the App Store. It made apps safe to download.

There was also conventional wisdom which had to be abandoned before the iPhone came out.

Nokia phones had 1 week battery life. They were extending it to one month.

The iPhone had a < 1 day battery life. Ultimately consumers decided that given the app experience available on the iPhone they could live with that.

This was a bet Nokia would have never made—-they were not in the user interface business. The fact that Apple—who was not in the phone business—-entered their market was bonkers. But once you make your own silicon anything is possible. Plus they had already done the iPod.

Well a founder can make decades long bets.

Any other company would have had an annual review of progress or a review any time mid-level leadership changed hands. Assuming a 50/50 chance of survival each time, the odds of the project lasting 10 years would be 1 in 1024.

Compute unified device architecture ?

I sometimes think about how good software quality would be if we all wrote test cases in the morning and coded in the afternoon. Motorola used to have a processes something like this…write and review tests at the same time as requirements are being approved, well before any coding starts. The number of defects we’d find in final system test was very small.

“Those who work with their minds must rest with their hands.”

Japanese woodworkers would spend the first half of the day sharpening their tools. The secret of Japanese woodworking is the steel they used…and their insistence on air dried straight grained softwoods. It helps make their precision joinery possible. It was also a technique designed for longevity so you could remove a piece from a hundreds year old temple, exactly replicate it and put a new one in. No metal fasteners meant no hidden weak spots due to rust which would fail in an earthquake.

Yeah nobody hears 16khz sine waves. But those high frequencies also make low frequency transients (eg square edges)sharper, and reduce ringing.

Chuck Norris does not use an assembler. He live codes machine language directly into DRAM.

He deal with security threats at the source, by tracking down the hacker and kicking them in the head.

Or there is only one game…a metaverse in which you create a new game by customizing the world through ai generated content and rules.

Fantasy below, Star Wars above—in a galaxy far far away.

They do art.

It’s just published on YouTube. Seriously the quality of diy videos and everything is sometimes PBS or BBC quality or better.

Because if you have a natural gas flame in air (lets say 2000C, my previous number was wrong as someone pointed out), then you can do useful work with it.

A carnot heat engine with a hot side of 2000C(Th=2273K) and a cold side of 200C(Tc=473K), has an efficiency of 1-Tc/Th or 79%.

So you can do useful work (hopefully inside the house, where it too will eventually end up as waste heat) AND recover energy from the exhaust gas of the carnot engine--which is 200C!

Imagine burning natural gas in your furnace, vs using your natural gas to run a electricity generator and mining bitcoin with it inside your bedroom. A Bitcoin rig makes a great space heater.

6) heated and cooled beds make sense for 8 hours a night.

A headache how? Because of noise or too much heat up above you?

A heat pump that is used on the exhaust of a low efficiency furnace might turn it into a high efficiency furnace. This would have other benefits because you would not have to throw away an old furnace for no reason.

ANY inefficient technology could get an efficiency boost just by using a heat pump.

But for this to happen HPs would have to be much much cheaper.

Another approach is to heat the interior walls and ceiling instead of heating the air.

Neglecting clothes for a few minutes, your body both radiates heat and absorbs it. When you are in thermal equilibrium you feel warm. If your walls and roof had low grade radiant heaters on them and emitted enough infrared to balance your own outgoing radiation, you’d be warm, even if the air temperature was cold and you had no clothes on.

So if you covered your ceiling with cheap thin iron sheet metal and built small induction coils behind them and heated the ceiling instantly, you could instantly feel warm in a cool room. Exit the room and you turn everything off…letting it lose its tiny amount of heat.

So proper system design could let you have low latency zone heating. Since you usually don’t use all the rooms in your house at the same time, there ought to be more than 50% efficiency in that sort of a system.

2. Our current approach of heating the whole house through forced air is much less satisfying than having a concentrated heat source (like a fireplace) that you can just walk towards and get as much heat as you want out of. Maybe psychological factors might play in…if you “could” be warm with just a little bit of effort maybe you won’t mind being colder than normal. Maybe heat gradients are better for you than uniformly high heat.

Could that have been solved by placing a tarp over the unit beforehand?

LBNL had a solid state thermoacoustic resonator the size of a coin, which I read about but never saw it in the market. It was supposed to be able to extract heat from anywhere.

That sentence is not aimed at the consumer, but is a very efficient mechanism for capturing government subsidies.

In theory burning high grade fuel to produce heat is stupid. CHP (combined heat and power) systems have more than 90% efficiency even with yesterday’s technology because all the inefficiency is used to heat your house.

The temperature of a natural gas flame is 2770C. Using it to heat a house to 30C is ridiculous. Instead you should run a gas turbine off of it to produce electricity. The exhaust should run a steam turbine which makes more electricity. Then a heat pump can capture more energy from the exhaust (and from your bath water etc) to pump more heat into your house. The electricity can be used to charge an electric car or run a gpu inside your house (and produce even more heat).

Heat pumps should be utilitarian and 10x cheaper, and used in multiple places. They should not be something you “invest” in with a 10 year payback time.

Pascal pointer notation was so logical. Dereferencing a pointer was just a caret to the right of the variable.

A double pointer was just two carets. And so on.

There was a struct symmetry about the whole thing. C broke that, especially with strict pointers.

Or the camera in the great pyramid…3 hours of opening a door, entering a tunnel only to find another door.

Veritasium is much better than Shakespeare. If you dropped his videos 500 years ago it would have caused a minor revolution.

This is a general problem. If you want to get a good answer to a certain class of problems (especially in the diy domain, and now even math problems), YouTube videos are the best. But you have to watch 20 minutes. This is not scalable.

Half a ton cargo capacity?

That’s like the weight of what, 2 fat people?

A ton is 1000 lb, right?

Well math is eternal.

Maybe printf will be eternal. But no window library is universal and eternal.

Perhaps we need eternal User interfaces and APIs to go along with them.

Edit: preface by stating that I agree with you.

Well I have a different approach. Sometimes I write and hack it to solve a particular problem. The code might be elegant or not, but if you understand the problem you can probably grok the code.

Next I generalize it a bit. Specific variables configurable parameters. Something that happened implicitly or with a single line of code gets handled by its own function. Now it’s general but makes much less sense at first because it’s no longer tied to one problem, but a whole set of problems. It’s a lot less teachable and certainly not self-evident any more.

The problem with math education is that we think the solution approach would be inherently superior to the first, and would make a better textbook—because it’s more generic. But that is not how real people learn—they would all “do” math the first way. By taking away the ability of the student to do the generalization themselves we are depriving them of the real pleasure of programming (or math).

Maybe back when paper was scarce this approach made sense but not any more.

Ideally I would love to present a million specific solutions and let them generalize themselves. That is exactly how we would train a ANN. Not be regurgitating the canned solution but by giving it all sorts of data and letting it generalize for itself. So why don’t we do this for human students? When it comes to education I think people have a blind spot towards how learning is actually done.

Education != learning.

Subtle difference. A parallel_for can have asynchronous threads. They can all divergen and run independent instructions (due to if statements etc)

SIMT means multiple processors all executing the exact same instruction. One instruction decoder, but like 64 execution pipelines.