I’ve got a textured PEI bed and when I’ve printed TPU, the adhesion has been perfect, i.e. good enough that the part wasn’t going to go anywhere unless I wanted it to, but still easy enough to remove when the print was done and the bed had cooled. I guess it could vary from filament brand to brand, so it’s possibly worth trying the same brand as I used, which was cheap Geeetech stuff. It’s £8 a roll, and I’ve used their cheap PLA for ages. I wouldn’t recommend their ABS+, though, as it seems to break down at the lowest temperature that gives reasonable layer adhesion.

So you think the reason a hygrometer can’t detect humidity ten minutes before it exists in order to start cooling the dehumidifier’s compressor to the temperature it needs to be to start working is human error?

I was going to share a graph from when I put a DHT20 hygrometer in my bathroom to prove to my family that the humidity was the cause of the mould and they should stop turning the dehumidifier off when its built-in hygrometer said it should be running, but unfortunately, it was long enough ago that Home Assistant decided I no longer need my one-every-ten-seconds readings and now only shows hourly readings, which aren’t enough to prove my point here. You’ll just have to take my word for it that when I did this test, I was surprised to find that although the humidity at the other end of the room started rising quickly after the shower was turned on, it peaked fifteen or twenty minutes after it was turned off again because diffusion without something like a fan or a draught moving the air around can be really slow.

My bathroom’s a weird shape as it’s long and thin and has a weirdly high ceiling at one end, so it’s not going to have typical airflow, but it is a real bathroom that really exists, and I did have data in the past showing it dried out faster if I manually turned the dehumidifier to maximum (so it would run even if its hygrometer said not to) ten minutes before turning the shower on than if I did it immediately before turning the shower on. Whether I’m going to shower in ten minutes is something I can know but a hygrometer can’t. This isn’t even really related to whether the dehumifier is smart as mine isn’t and I can operate its switch as easily as I could operate a smart switch, and my shower isn’t electric, so there isn’t a switch I need to operate before using it that could be made to do two jobs

Again, I’m not sure what components were used in the older model, but given the age I’d be very surprised if the electronics it uses would be more expensive to manufacture than the newer one.

That’s fundamentally where you’re going wrong, then. 1980s electronics (for a dehumidifier, it wouldn’t even be electronics, it’d be electromechanical) are often much more expensive than modern approaches, and even when they’re cheaper, it’s typically not by much. Over time, it’s got cheaper and cheaper to precisely make small things, but the costs of materials haven’t meaningfully gone down, so the 1980s approach costs about the same as it did back then, whereas digital electronics have plummeted in cost. Now, anything where the best approach was electromechanical rather than electronic is almost certainly cheaper to do digitally.

Take toasters for example, most toasters don’t have a timer at all. They have a little piece of metal almost touching a contact. When you turn the toaster on, that metal heats up and it bends until it touches that contact, ding toast is done.

Another great example of being out of date. Fifteen years ago, cheap toasters almost always used a bimetallic strip and the dial controlled the position of the contact it touched so it would have to bend more or less before it disconnected. In nearly every modern toaster, however, you’ll either have something like a 555 timer and the dial will control a variable capacitor that changes the frequency of an oscillator to make it count slower or faster, or it’ll have a dedicated toaster control chip like the BCT5512 and the dial will control a potentiometer that a capacitor drains through. Mouser list the PT8A2511PE toaster controller for £0.111 in bulk, but the cheapest bimetallic switch they carry (which is too basic for a toaster because it’s got a fixed switching temperature) is the F13A17005L360100, which is £1.93 in bulk, more than seventeen times the price. (I suspect they used to have cheaper ones back when toasters still used them, and they’ve been discontinued now toaster manufacturers have stopped ordering them.)

But it’s cheaper and simpler to just do it the old way, and for many applications, that’s fine.

A lot of the time, the old way is more complicated and more expensive. Technology doesn’t just let us do things we couldn’t before, it also lets us do existing things in new, better ways, and being cheap is one of the most in-demand things. It’s lower tech to hire ten labourers with shovels for a week to dig a hole, but it’s much cheaper and faster to hire one labourer with a digger to dig it in an hour.

Hell, I’m certain there are dehumidifiers on the market that don’t have any kind of humidity sensor at all. Even simpler…

Having no sensor at all is certainly the cheapest way to do it, but we were talking about ones that do have a sensor, and whether, once you’ve opted to have a sensor, there’s any major cost to making the device smart. If you’re aiming so low-end that you don’t even have a sensor, then you’re clearly not concerned about the marketing benefit of extra features, so wouldn’t bother making it smart.

The cheapest hygrometers these days only have a digital output, and a wire and a potentiometer aren’t going to be able to query an i2c bus to ask the hygrometer what it’s measured without the help of a microcontroller (and the microcontroller might be cheaper than the potentiometer anyway depending on the specific model of each - have you actually looked at the 2025 prices of things before making assertions about what they cost?). The analogue component of a hygrometer that actually does the measurement gives fairly small changes to the resistance/capacitance (depending on the kind of hygrometer), so the results need amplifying. If you’re measuring on the same chip, you can get away with a simpler amplifier and digitally compensate for any nonlinearity, whereas to get a strong enough signal to make it to the rest of an analogue circuit without much degradation, you’d need an amplifier that ends up being more complicated than doing everything digitally.

You’re missing my point. It’s likely that the cheapest way to design and build a dehumidifier these days will already include a microcontroller interpreting results from a digital hygrometer because these components are cheap and easier to work with than purely electronic/electromechanical designs with no microcontroller. The cost of switching from a non-WiFi/Bluetooth/Zigbee microcontroller to one with one or more of these networks is negligible, and once you’ve got it, it’s not meaningfully more expensive to pay a software engineer to expose the on/off switch and hygrometer readings via that network and have the marketing people write Smart! Now with WiFi! than it is to skip it and pay the marketing people to come up with some other nonsense to put on the box. If you care about security as little as the average IoT vendor does, then it’s nearly free to turn a dumb device into a smart one, so if it makes a handful of extra people buy the device, manufacturers will make things smart. For a dehumidifier, there are reasons why a handful of people will prefer a smart one, so smart dehumidifiers get made.

In the UK, it’s common for electric showers to be on a separate isolator that needs to be turned on before they’ll heat up, and it also activates an extractor fan, and most people turn it off again when they’re done showering. It’s pretty simple for a home automation hobbyist to swap the regular isolator switch for a smart one, and then their system can know when they’re about to shower and activate the dehumidifier immediately. This can be much better than waiting ten minutes for enough humidity to diffuse into the dehumidifier for the humidistat to activate then waiting another ten minutes for the cold side to cool enough for any dehumidification to start.

I didn’t say a home automation system would be measuring the humidity and reacting. The opportunity to do better comes from the potential to be more proactive if you can figure out a way to tell a computer about impending humidity.

Don’t give JK Rowling ideas.

I can get a board from AliExpress with an ESP32-C3 on it with free shipping for £1.10, so I’m not inclined to believe the £0.765 unit cost for a 5000-part reel from Mouser is really the cheapest way to get them in bulk as the other parts on the same board and the shipping have to cost something.

The cheapest hygrometer that Mouser sell is £0.748 per unit for a 10,000-part reel, and its datasheet says not to leave it for more than 60 hours in greater than 80% relative humidity (which is a pretty likely scenario for a dehumidifier) as it’ll drift, and if it happens often, it’ll age faster. You need to spend more to get rid of that restriction. I’ll concede that the accuracy penalty if you cheap out isn’t as bad as I thought - I’d not actually looked at a datasheet to see how badly modern hygrometers would drift, I just knew that they did - so plenty of manufacturers wouldn’t care, but the parts are still comparable prices, not a factor of ten like you’re claiming.

Originally, that was from Dave, which is owned by the BBC.

The built-in hygrometer’s not necessarily going to be as good as a well-designed home automation system, especially if the fan’s not running all the time, so it has to wait for damp air to diffuse into the machine. It also lets you do other things, like not bother turning the dehumidifier on if there are open windows if you’ve got some way to detect that, or report the humidity to something that will graph it. It’s not stuff that most consumers will care about, but a microcontroller with WiFi like the ESP8266 or ESP32-C3 costs less than an accurate hygrometer chip, so it doesn’t make much, if any, difference to the final price, particularly if the product was going to use a microcontroller anyway.

Sveriges Riksbank Prize in Economic Sciences in Memory of Alfred Nobel winning economist. There’s no such thing as a Nobel Prize In Economics, and economists were upset by this and made their own prize with a complicated name that the media would shorten and muddle with a real Nobel Prize.

The same site says things like:

Between 1901 and 2024, the Nobel Prizes and the Sveriges Riksbank Prize in Economic Sciences in Memory of Alfred Nobel were awarded 627 times to 1,012 people and organisations.

which pretty clearly makes a distinction between the Nobel Prizes and the Sveriges Riksbank Prize in Economic Sciences in Memory of Alfred Nobel.

There’s no such thing as a Nobel Prize in economics. Economists got salty about this and came up with the Sveriges Riksbank Prize in Economic Sciences in Memory of Alfred Nobel, and rely on the media shortening it to something that gets confused with real Nobel Prizes.

If it works like real WoW64, then 16 bit applications won’t work ever but 32 bit applications that don’t work will be because of fixable bugs.

That way they can sell it as Crab Product rather than the much less appetising Crab Flavour Fish Product.

Depending on the era of the game, you might well own a copy of a game on a disk, just like you own a copy of a book when you buy a book. Weaselling out of first-sale-doctrine stuff came a long time after people started buying video games. A century ago, publishers were trying exactly the same thing with books, and depending on the country, either legislation was introduced that made it explicitly illegal, or the courts determined that putting a licence agreement in a book just meant that the customer got a copy of a licence agreement with their book, not that they were bound by its terms.

You can’t make an LLM only reference the data it’s summarising. Everything an LLM outputs is a collage of text and patterns from its original training data, and it’s choosing whatever piece of that data seems most likely given the existing text in its context window. If there’s not a huge corpus of training data, it won’t have a model of English and won’t know how to summarise text, and even restricting the training data to medical notes will stop mean it’s potentially going to hallucinate something from someone else’s medical notes that’s commonly associated with things in the current patient’s notes, or it’s going to potentially leave out something from the current patient’s notes that’s very rare or totally absent from its training data.

Sandwich cake is already a term that means the same thing as layer cake. The classic combination of two layers of Victoria sponge with strawberry jam and whipped cream in between is called a Victoria Sandwich. Anyone arguing that a layer cake isn’t a sandwich is just illiterate, not a defender of semantic specificity.

AGPL is a full-on FOSS licence with strong copyleft requirements, not a measly open-source licence like Apache, which could be pivoted to proprietary at a moment’s notice. We’re communicating through an AGPL-licensed system right now as it’s what Lemmy’s licensed as. If they were going for a corporate-friendly licence, AGPL is the last thing they’d choose as it forces you to share source code with even more people than the regular GPL does.