I've never hit anything with Qwen. Very rarely it will silently switches from 3.8 to 3.7 (which makes it notably dumber) but you just click the drop down at the top to switch back. It's rare though. I have >week convos with no switching

The issue is that there is ultimately a limited market for having access to space. SpaceX's business model is premised that if you make access cheap enough that there will be a market for multiple launches a day. But to me demand seems very finite? You can make it 10x cheaper or 50x and it sort of doesn't matter.

There just isn't much to do in space. You have:

- Telecommunications.. Can scale only so much. Starlink is cool. Compared to Iridium it's way way cheaper. But demand is .. modest? It seems to only beats fiber on the ground in special situations

- Telescopes looking up. This is limited by science budgets that aren't going to change much.

- Telescopes looking down. Those people have infinite budgets and already have all the satellites they'd want. If you're not in NATO then can't depend on SpaceX anyway - so they're not capturing that market.

- Tourism. Vomiting for several hours in zero-G will be a thrill for some for a bit, but I don't see this being a huge industry. Maybe on par with basejumping or something.

- Mining.. seems impractical scifi.. premised on wanting to build huge structures in space - but again, for what goal ultimately?

It's sort of like if you made access to Antarctica 100x cheaper. It'd be great, but it's hard to imagine it's going to suddenly open up a ton of demand.

But I'd love to hear counter arguments :)

Indeed SpaceX is boiling multiple oceans at the same time and betting the timing will align.

-- SpaceX is building a starship factory, not just starship. The market won't deliver enough demand, they need to create their own

-- V3 Starlink constellation targets 100k sats. With 50 sats per flight (100t to orbit) you need 2000 flights to build constellation and then 400 flights per year to replace sats after 5y lifespan.

-- There was that idea of ballistic human travel but this is unlikely to materialize

-- There also was solarsat/powersat where solar power is beamed from space and if you massaged your math the right way the numbers were ~promising

-- AI saves the day. We "need" 1mln AI sats. That solves the demand and justifies the spending. for now

the demand for bandwidth is infinite… this used to mean copper -> fibre, but it’s that squared for orbital bandwidth, as they lower and close up the sat formation, the antenna gets smaller and the data rate gets faster … I think Musk killing all terrestrial telcos is quite likely

> the demand for bandwidth is infinite

Is it really? Our brains have a limited bandwidth, and there's no need of more than a small multiple of that per person.

> Telecommunications.. Can scale only so much. Starlink is cool. Compared to Iridium it's way way cheaper. But demand is .. modest? It seems to only beats fiber on the ground in special situations

It is not designed to compete with finer. Go somewhere truly remote and see how it’s changing the world very quickly. I’ve seen it first hand in Arctic Canada, Australia and North Africa. Before starlink the best they had was dialup. Now they video chat family and doctors, do school online and so much more. Easily a few billion people will use it.

You also forgot about AI data centres in orbit, on which SpaceX have pegged virtually all their valuation in.

Theoretically couldn't you take an electric car and put an off-the-shelf generator in your frunk and wire it up to charge the battery as you drive?

My PHEV won't 'start' while plugged in. You'd need to bypass the usual charger mechanisms, and then the computer seems likely to get mad...

A minor inconvenience compared to the thousands of charger cables and wall boxes that would charge driving the cars for a very small moment.

>>and put an off-the-shelf generator

Even my really underpowered PHEV has a 65kW electric motor. No "off the shelf generator" can generate 65kW, in fact it would be a huge unit that could.

We could of course say - well, it doesn't need to match peak usage, it just has to charge the battery faster than it depletes - you still probably need around 20kW for steady motorway driving, and that's a pretty hefty generator.

I don't think you need to have it match the depletion rate. Even at say 10kW you'd doubling the effective range. The point is people want to be able to use their car for that twice a year road trip, or drive to the inlaws for Thanksgiving. If you can attach something that extends the range for these occasion then that's probably enough.

Furthermore, the generator can run 24/7, but realistically you will at most drive ~6 hours a day. So for a leisurely road-trip this may be enough to avoid needing stops at charging stations

20kw would be enough to maintain the battery though? So you could halve that and 'only' double the range in the worst case.

JavaFX is supported by Gluon and not Oracle

It seems more Oracle than Gluon: https://github.com/openjdk/jfx/graphs/contributors

Oh hmm, I'm honestly not super sure then. I know Gluon is the one that's seemingly has a business providing JavaFX support, while Oracle.. as far as I know doesn't offer a license or anything services in that direction? But I could be wrong. Gluon is mentioned in the article.. they have a very Qt-style confusing licensing situation where it's probably all free.. but you're never quite sure :)

The whole space is a lot more icky than one would like it to be.. Which is unfortunate b/c while it doesn't have a billion features, it's a nice composable reactive framework to use (at least from cljfx/clojure)

It is complicated.

There was a time Oracle decided it wasn't worth the investment, note that it started as E3 scripting under Sun and it was only finalised after the acquisition.

So a few Java champions took over it at Gluon.

Then when Java 26 was released, Oracle reintroduced support for JavaFX,

https://www.oracle.com/news/announcement/oracle-releases-jav...

Oh huh, I guess it's pretty recent news. I wonder if it'll lead to new features. It's nice to see people are working on Native builds for instance, and that Graal integration is now an official priority of sorts.

It’s all a circle jerk in the JVM space

Thanks for writing this up! It still assumes a lot of prior knowledge, but it's a good starting point to approaching these problems. How JavaFX works on different platforms really looks like dark magic from the outside. I wrote an app in Clojure using cljfx. I can jpackage it and run it Linux/Windows/macOS .. it's quite easy to get working. But once you step outside of the standard platforms it gets complicated really fast. Trying to get it to Android or a native build and suddenly you have to be an expert in several build systems and JVM internals.

is there a good metric of model degredation over time?

Im a bit lazy and only use the free models different companies host and the biggest difference i see is that some models (Gemini, OpenAI) get progressively stupid in long chats. You end up having to start a new session every oncr in a while. Or they get really hung up on a theme and cant shift to a new topic.

By contrast, Ive been impressed with Qwen. I have some chats on research and code architecture that have stretched for weeks without any noteable change in quality (though occassionally it seems to "rush" to an answer)

Im just looking at all the listed benchmarks and im unsue which i should be looking at

Isn't HarmonyOS a fork of Android? I've never seen anything about the apps being half broken - do you have more info?

The newer versions of are no longer based on Android. So it's either quirky VMs with hundreds of papercuts or nothing.

Huawei Global is still a thing, and in some markets they keep releasing phones using Android 12 and no GMS. That is probably even worse.

Seems this would make it possible to make a relatively cheap waterproof actuator if you made it out of something corrosion resistant?

I've been looking for actuators to use in damp/wet environments and sealed actuators are rather expensive and supposedly don't last too long

Some hot water radiator systems use wax actuated valves made from brass (Taco brand is common). They are extremely simple and robust.

thank you for the pointer!

I've struggled to protect eletronics. Waterproof enclosures all fail over time. Though I have some ideas for immersion in mineral oil. But finding actuators that could work has been a bit of a struggle. These look cool, though maybe not something one could run off 5V :) Ill keep looking around

This is why wax motors are used for washing machine door locks.

how does it work for a door lock?it seem it works pretty slowly.

Mayne you know, are the internal components already designed with corrosion in mind?

In washing machine application the locking action is quite fast as the heating element is PTC connected more or less directly to mains, pretty substantial amount of energy gets dumped into the wax over few AC line cycles. In that application (and also in most industrial automation situations) ~100ms is more than fast enough because when most of the effector components are AC powered you end up with ~40ms worst case actuation delay just from the physics of AC electricity. On my washing machine it is quite noticeable that when you press "start" the door gets positively locked well before the capacitive sensing "button" finishes playing its haptic feedback non-sense. (the fact that front panel of cheap-ish toploading washing machine is actually a huge touchpad with haptic feedback that is designed to look and behave like five physical buttons says many things about economics of manufacturing stuff for that somewhat hostile environment at these scales)

oh interesting - I didn't realize PTCs can be used that way safely. I thought it hits a target temperature and then will continue heating at that temperature (which in your scenario would be unsafe). Maybe for my usecase it'll be hard to find a suitable actuator b/c I have a 5V system running of battery (looks like 12V ones do exist though)

From the linked page:

> When a cycle is started, a wax motor is actuated pushing a pin outward and locking the door. This design has cost, reliability and safety advantages. In moist conditions a wax motor costs less for equivalent reliability than an electromagnetic solenoid or motor latch. It has a predictable passive release delay. If power is lost the door remains briefly locked, designed to be longer than the high speed spin cycle coast-down time, then reliably unlocks as the wax cools.

furthermore, couldnt you do parallax from the multiple cameras as well as flicker the flash?

seems pretty easy to make it sufficiently difficult to trick the system

There's very little incentive to switch. When people release for Linux they almost always first/only test it works on Ubuntu - hence it's going to be the least buggy. (ex: GOG only seemingly tested games on Ubuntu)

So the switching cost is buggy software. You'd need something radically different that brings enough new features to the table to make it worth it the switch and dealing with bad/non-existant support.

I think something like Nix/Guix but with stable library versions that matches Ubuntu/Debian LTS ones 1-to-1 could get traction

Then you would have to embrace their bad decisions, such as this one, and if you do, what’s the point?

Everything just needs to run from a container with their expected runtime environment.