I did something with my Bosch washing machine (not like the OP). My washing machine is at the other end of the house from my home office. Sometimes I would put a load of washing on, and despite setting an alarm, might forget (perhaps I am in an important meeting etc).
So I decided to solve it.
Using the Bosch API - I can tell both when a cycle is complete, and if the door is open. Currently I use their default version, but there is a local hosted option I'll be switching too now the proof of concept works.
So using Home Assistant I have a simple script that detects when a washing machine cycle is complete AND the door has NOT been opened. This implies my washing machine has wet clothes still in it.
So Home Assistant will alert my phone (and my wife only if she is home based upon presence detection) once every 15mins that there are wet clothes waiting in the washing machine.
My washing machine is a "dumb" machine from the '90s. The wash cycles run based on the position of a glorified timer knob: it doesn't have a computer or sensors to detect if it needs extra time aside from the water fill stages. Thanks to this consistency, I just set a 40 minute timer on my phone, and it's always done by then. Can't get much simpler than that. If I need reminders, there's always the alarm snooze function.
It's actually really, really easy to get the state of a "dumb" washing machine (or any other electric machine) into Home Assistant using a smart plug. You can use something really basic like "power draw for > n seconds followed by no power for > m seconds" to detect when a cycle is finished. You can get way fancier and look at power draw curves to determine what part of a cycle it's in, or which cycle, if you really want to. You can add a door sensor (recommend Samsung) if you want to know if the door's been opened.
Unfortunately it's much harder to do the same for an electric dryer, since there's no inexpensive or good smart plugs for 240V last I checked.
For the dryer think even simpler: get a Esp32 with a SW-420 vibration sensor, stick it on top of the machine somewhere, and look for vibrations. Since most dryers don’t have cycles, it’s really easy to detect if they’re running or not. There’s even a potentiometer on the SW-420 so you can adjust the sensitivity without reprogramming the ESP32.
This is (largely) what I have for my "dumb" washer. What makes mine unique is the washer is in my apartment building's cellar, too far for Wifi. So I have a LoRa transmitter that routes the message to a LoRa receiver on my home network.
Unfortunately, as you noted, I haven't figured out how to handle the dryer as the load is too much for all the smart plugs I've found. I wish there was a clamp-style monitor on an extension cord, but it seems that is something which doesn't exist.
Thankfully, the dryer isn't as much of an issue since dry clothes can sit there until I remember to get them.
I wonder if it would be practical to McGyver a solution into place by coiling some wire around / next to the dryer's power cable - to detect flowing current - and connecting the coil to an Arduino / Raspberry Pi
My understanding is that detecting an electrical current through a power cord typically doesn't work when measuring all the wires together, because the current in the live wire is canceled out by the return current in the neutral wire. This is why clamp-style meters are usually put at the electrical panel, where individual wires are accessible. Since I live in an apartment, that isn't an option for me.
It's also possible that your machine lets off some elecromagnetic noise you can detect directly.
The other thing you could try is put some sort of a microphone / vibration on it, then train a small NN on the waveform to distinguish between the washer and dryer.
I considered that, but the washer and dryer are stacked (limited space) so the vibrations of the washer would be picked up in a dryer vibration sensor.
Is that really a problem though? If you're washing another load at the same time as the dryer is running, wouldn't you generally want to wait until both are done anyways so you can swap the wash to the dryer after getting the dry load out?
Also at least in my experience the dryer takes longer to run than the washer, so if I've just rotated a previous load and started both at the same time I'm always waiting on the dryer to be able to do the next one.
Not always. I only put sheets and towels in the dryer, everything else gets hung up. So I may do two or even three washing loads in the time of a single dryer load (it can be very slow). So I'm much more interested in when the washer finishes than the dryer.
You might be able to set something up using a current clamp connected to an esp32, though it might have to go over a point where the wires are separated so possibly inside the machine. Might not be too hard depending on how you feel about opening up your dryer.
I know nothing about US 240V power circuits — what plugs do you use, could you get by with a Euro system?
I use Eve Energy smart plugs, which seems to be supported in Home Assistant through the matter integration. Local first, no bullshit remote account requirements, good quality, around 40€ / USD 45.
> I believe that US 240V is 2 hots, neutral, and ground.
Correct. Its called split phase, a 240V transformer is center tapped and that tap is grounded to create the neutral. Either end of the transformer to neutral is 120v and end to end is 240V.
> EU 400V (380V) is 3 hots, neutral and ground.
Three phase 230/400: 400V is line (hot) to line with 230V line to neutral. More for industrial use but I hear some homes can have this service for whatever reason.
> None of this is cross-compatible.
Not really. A 230/400 volt system also supplies 230v single phase. A 230 volt European device will work fine on 240V split phase unless it has a motor which will run faster on 60 Hz which could overload it. Though I have a machine with a three phase 380 volt 50 Hz motor running happily on 230/400 60 Hz from an autotransformer supplied by 120/208 60 Hz. Just runs 20% faster.
> Three phase 230/400: 400V is line (hot) to line with 230V line to neutral. More for industrial use but I hear some homes can have this service for whatever reason.
It varies from country to country in Europe. In the UK you'll almost never find 3-phase in a home, in Sweden even apartments usually have 3-phase supply. In my Swedish apartment the only thing connected to more than one phase is the induction hob.
Extra credit for discerning washer vs dryer, recognizing "done" buzzer/bell, etc. Might be useful if your dryer has a "wrinkle guard" feature; buzz and then keep tumbling on low heat.
For the washer, you could probably also discern portion of the cycle (fill, agitate, spin, drain).
I do this for my washing machine, drier, dishwasher and also 3D printer, in the UK. Have some Z-Wave Greenwave PowerNodes, hooked into Home Assistant. Works great. I also have Alexa hooked into Home Assistant so I was able to make it announce through our Echo Dot when a cycle for any of those devices completes.
I use an AI agent running on an old Android phone to monitor the position of the switch with the camera and let me know when it's straight up and down or side to side, since those are the four off positions of my knob. (kidding but maybe not by this weekend)
Keep that dumb washing machine from the 90s, I can almost guarantee you that a new washing machine is not meant to last as long. Maybe 6-7 years if you are lucky.
The article says that a washing machine used to last 20 years, and now only lasts 10 years. However, it also says that machine usage has doubled, from four to eight times a week. So, the new machine lasts the same number of cycles, but the number of cycles is reached much faster.
An article from 2025 is pulling data from a 2003 study that compares usage from the 1960's to what I assume is 2003, right when He and CPU controls were starting to become popular.
My 1990's Whirlpool lasted for 25 years before a barrel support rusted out. It had a mechanical timer and did a good job cleaning my clothes. Now I have this awful HE LG thing that I ALWAYS use the bedding mode otherwise it just mashes the shit out of everything destroying my clothes. HE feels like a conspiracy by big clothing to sell you more clothes, destroying the environment while fooling you into thinking you're saving it by conserving water.
There's an easy fix for that, put your clothes in these 'washing bags' that was always good to prevent wear but sounds like that might be what you need
Mashing the shit out of it is what actually washes your clothing without high temperature and minimal amount of detergent. I can't stand top loader washers that are ubiquitous here in NZ - they are so useless.
We’ve had a few scares. Last year, it didn’t stop filling and caused some flooding across a few rooms before we caught it. Earlier this year, the plastic on the knob mount was so brittle it snapped off. We’ve been servicing it, and we’ll keep it going as long as we can.
Miele from my grandmother still running here, 1988 machine! No defects, still pristine inside.
No guarantee though that today's Miele is the same quality.
They are. Bought a washing machine more than a decade ago, still going strong, and a dishwasher more recently and it was every bit as solid as the rest.
But the fast cycles of olden day cames at a price in terms of power draw and water usage, so it’s unfortunately similar to missing incandescent light bulbs.
I occasionally do a washing load before bed that I know I might not wake up for to put into the dryer. Fortunately, my machine has an "extended tumble" cycle of sorts that will keep the clothes fresh all night at the expense of a bit more water, but while saving my bedtime routine. We end up with a lot of these nighttime loads because we're toasted all day watching kids and we prioritize laundry off-peak electricity hours. Love my Electrolux, but I imagine many brands have a comparable feature.
The equivalent on mine (a Bosch) is to wait to start anything at all until 1 cycle-time less than ten specified number of hours. Churning all night instead seems a peculiar design choice.
Does your machine not have a delayed start function? I’m standing in front of about 40 washing machines right now and they all appear to have this function.
I have been planning to implement something similar with my countertop oven - however having no API or other connectivity, I was planning on simply plugging it via a smart plug, and using the power draw measurements to determine whether it's idle or not (that is, arm when power draw transitions to above idle, then alert once it drops back to idle).
Yeah I tried to use the builtin sensor on my LG one but it turns out, there's no 'door open' sensor per se, only the 'locked successfully' signal. So I had to add an external Zigbee reed switch door sensor..
I have a G-Shock 5600 watch that can alert me when my washing machine is finished. At the start of the cycle I take note of the total time it takes, I set that time on the timer of the watch and hit start. It will beep once the washing machine is finished. Been doing that for about 15 years now.
a) Australia is a very large, geographically diverse country that has committed to rolling out fibre in rural areas. It is an example that it can be done and you don’t need to rely on bandwidth limited satellites.
b) Our internet infrastructure was sub-optimal but is being upgraded to full fibre to the door across the country. 1g is available to consumers today, 10g to businesses, 100g in testing. And it’s future proof.
Excluding the part of how the access was gained (i.e. fault of end user etc) - I would have thought Stripe would have far better security measures in place for obvious signs of fraud?
If someone changes the bank account of a business for funds to be sent to - why not have a short (?48 hours) delay in allowing funds to be sent there. Especially if the business has been running a while (not sure in this specific case) with no bank account changes - then its not unreasonable for a built in delay for new accounts to be used.
Coupled with some basic stuff of "new bank account, new express accounts, spike in charges, sudden request(s) for withdraws" seems like something that a smart team in Stripe with their development experience could easily implement as an automated security trigger, and hold the funds pending additional security checks?
If it’s $200 in parts then a premade item would be at least $400 and probably $500. There probably isn’t much of a market at those prices.
The bill of materials needs to come way down to sell it for much less than that. Buying the parts in hobbyist-level bulk (100 pcs) probably would only shave $100 from the final price.
I had dreamed of physical "analog" controls as a standard feature on keyboards and input devices for various applications -- would be great productivity booster for power users.
Nice - be interesting to see what kind of integrations they have for the knob (at a bare minimum I better be able to play break out and scrub video in da Vinci).
I use a Keychron Q2 which has a dial at the top-right - it's mostly pretty good though I wish it were a little bit easier to program and of course it doesn't have haptic feedback.
While I wish I had the time and talent to pursue this as a hobby - and I dabbled in PCBs and DIY electronics in youth - life has other plans.
I see some sibling commenta that link to keyboards with a knobs-- that is cool - but I am imagining having access to more than one knob so I have a bank of at least 5-6 knobs, slightly smaller than the DIY project, so I can program them for multiple uses like we do for Function Keys today.
Imagine stream deck with knobs instead of buttons - and available either as a separate USB/wireless device or as a simple "nuneric keypad" / "function keys /media keys" kind of addition to keyboards
I made a SpaceMouse the other day, it was really easy, mostly just the print. It took around an hour to assemble everything, I think? It also worked very well.
Ah yeah, that's much more work and I don't think it uses the official driver, which does a lot of the heavy lifting. I think maybe start with the other one and see how you like it, since it's mostly a few minutes of soldering.
"They might not have "known", but come on, you're selling radiation-hardened chips to NASA. "
But do people ever actually "invoice NASA" for components. It was probably one of 100 different sub contractors building the actual circuits to NASA specifications, i.e. it was lower in the chain rather than NASA itself.
(Doesnt excuse the non-disclosure to those subcontractors)
>But do people ever actually "invoice NASA" for components
Yes, absolutely they do. I'm not a part of this mission, but I'm currently working on another NASA spacecraft mission. I don't know the percentages off hand, but a substantial portion of our spacecraft is built in house with parts purchased directly by NASA from the manufacturer.
Regardless, there are lines of communication to subcontractors. The mere fact that they found out about this at a conference is significant evidence that Infineon didn't notify who they should have.
Off-topic, but when components are sourced directly from the manufacturer do you have to buy in bulk? I figured you didn't just go on Mouser or DigiKey, but I would think manufacturers don't like dealing in small amounts.
For spacecraft parts, they absolutely don't mind (they're charging for the privilege of course). For the parts I'm familiar with, we generally buy both the necessary flight-rated components (both enough to build the vehicle and some number of spares) and a number of unrated components used in various test apparatuses in a single order. Once you get down to the level of stuff that's not even a flight-test fixture, we can indeed source parts from pretty much wherever. The biggest issue then become US government procurement rules that require us to buy American, but I'm pretty sure I've seen at least Mouser get used before.
I'm not a machine learning person - so I'm confused about this.
As someone who doesnt understand ML - I have always assumed the whole point of ML is to try different things in the game, almost randomly, and over (long) periods of time the AI gets better and better at the game.
If having a single unexpected event causes such a large swing in outcome, and the AI cant "explain" what is different to cause the swing, then what exactly is the ML doing for it to fail on such a seemingly simple change? Doesnt that defeat the whole purpose of this?
I'm obviously missing something obvious - because I would assume the real goal of ML is that it can teach itself the game, even if that involves unexpected situations, as a human does?
This article doesn't describe it in detail. One scenario imaginable would be that they ran their model trained on non-full moon data for an evaluation on a full moon day. Which means the model would simply apply it's learned "optimal" action policy in a different environment, where the previously learned action policy doesn't lead to good scores anymore.
So does this mean if they allowed the game to run on "full moon days", it would be expected to eventually get a higher score (if the full moon day allowed that through the actual game mechanism)?
Yes, the full moon day can help you get a higher score due to the bonus luck you get that day. On the other hand, the full moon day makes werewolves (and wererats and werejackals) a lot more dangerous because they'll always be in animal form.
When you try to fight a werecreature in animal form it can summon large numbers of animals of its kind to attack you. This can be extremely deadly for a player who is unaware of this ability. An experienced player knows to attack werecreatures only at range or avoid fighting them altogether. However, encountering the werecreature in its human form is much less dangerous unless it's carrying a powerful weapon.
It's not a single event, it's more like a new general game state that was never seen during training. Imagine learning to play the violin really well and then someone changes the way acoustics permeate. It doesn't matter if you're a human or an ML algorithm, you're going to have a hard time playing like before.
But something is wrong in the learning, because as a human NetHack player who has ascended, I can say that we don't play radically different on full moons. Yes, the random numbers go your way slightly more, but that's about it.
This tells me the algo is trying to hard to predict the game or learn a decent static strategy, rather than make situational decisions.
The issue is with werecreatures on a full moon. Most humans exposed to Western culture (likely all NetHack players) have heard of werewolves. I think it’s safe to say that everyone who has heard of werewolves knows they are most dangerous on a full moon. Even if you are a total NetHack beginner you know to avoid these monsters on the full moon. The game even helpfully reminds you of this information both by telling you about the full moon and by having a werewolf howl incessantly when it’s on the same level. However the game does not explicitly fill in the gap for you. It expects culture to do that.
The advantage of human common sense over machine learning models — at least when it comes to role-playing games — is that we carry around a ton of this cultural information. A model trained only on NetHack — not on broader culture or folklore/fairytales/mythology/fantasy — is simply not going to be aware of this link between full moons and specific monsters becoming more dangerous. So if it’s developed a fairly naïve strategy of just fighting or avoiding everything in its path based on a model of relative strength then it’s going to be tripped up when an outside event (the phase of the moon) upends that model.
I disagree. There are a lot of idiosyncrasies about monsters in NetHack such that getting good at NetHack is 99.9% about learning the NetHack world, not the real world. The werecreature game mechanic is no different than the POI or HALLU effect, so I don't think the AI needs any special knowledge. I
bet it comes down to how much memory the algorithm has, since the transformation might occur way later than being bitten, while most poison kills are fairly quick. The problem is NetHack requires you to have at minimum 1000 turns of memory to know when to pray. Even more if you want to keep track of where stuff was.
Tfa states that the agent was trained for points, and an other user states that some critters are a lot more dangerous during full moons.
Wouldn’t be very surprising if the agent hyper-optimised farming those critters for points. It would not be able to change strategy if the cost/benefit of that farming changed massively, so would now be performing significantly worse.
Humans train simultaneously as they operate, and humans can see the message about the full moon.
If nobody includes the full moon message as input to the ML model, and tries to operate the ML model with the training it has achieved running in non-full-moon mode, its operating score in full-moon-mode may be lower.
Even if it had proportional training time against full-moon-mode to incorporate that into the model, if you don't tell it when full-moon-mode is active wouldn't the optimal behavior be to optimize the score for 27/28 days vs 1/28 days of the month?
If full-moon-mode is an input to the model, then it can trained to optimize for both scenarios.
So for ML to work, it has to know all permutations of a game? Does that mean ML is useless for non-deterministic games with random outcomes or procedural generation?
The thing it has learned previously needs to apply to the next run.
If you train an ML model on thousands of attempts at going around some racetracks where touching the walls slows you down, and the score is achieved by executing a fast lap, and the inputs to the model include where the car is and where the walls are, it should optimize towards avoiding touching the wall.
This behavior would likely still work even on new procedurally generated tracks that the model had not previously seen, as long as the relationship of inputs (car, walls) to desired behavior (fast lap) still applied.
If every N number of runs for a large value of N the game changes so that the walls are actually speed boosts and the center of the track slows you down, and there is no input to the ML model to tell it that the situation is different, it will initially try the previous strategy and perform worse, and it will be difficult to train it to handle both versions of the game without some discriminating input value to train on.
Well that is the whole trick. ML models ideally generalize from the training inputs to whatever new inputs show up during inference. For example, a vision model should recognize an image of a dog as a dog even if that exact image was not trained on. But that generalization always has limits. Usually score will decrease substantially the further "out-of-domain" the inputs are. So this model works fine when running a randomly generated dungeon it has never seen, but not when running a set of game rules it has never seen.
I'm trying to find games for my kids that would have the same influence. Fortnite etc is all so popular - but I feel that Police Quest, Hero Quest etc are a big part of my logical reasoning skills I have today...
I found a few good Android games for my kids (god it's hard to get through all the ad-ridden garbage)
* Monument Valley
* I Love Hue
* Battle of Polytopia
* Grand Mountain Adventure
* Tiny Bubbles
* Kingdom Rush
* Planes Control
* Human Resource Machine
No quests however, so I would also welcome suggestions here
For real, kids games are absolute garbage. I would pay for a curated list of games that are engaging, require some amount of thought, but that are not massively stuffed with ads.
What I find strange about this is you dont need it to be "deepfake".
Just an inside job.
If a large company allows a single employee to transfer millions to a new bank account/vendor that has no history, on "their belief" the instruction came from an approved person (i.e. their boss, CFO etc) - that company has major governance issues that are not related to deepfake.
Imagine the more simple scenario - an employee transfers millions, knowingly fraudulantly, to some people they are working with. They then simply supply some "deep fake" pictures and a story how it was an accident - and boom; you walk away with millions.
Checks and balances exist for many reasons - deepfake doesnt overcome those by itself. This company is just missing basic steps that would have protected itself here.
edit: in fact- its even more obviously some inside job; put the deepfake aside for a moment. How was the meeting even booked? Their PR person said "none of our internal systems were compromised". So this meeting magically appeared in someone's calendar? Using their internal video system (Skype or Teams or whatever). And the criminals knew to target this person, with enough knowledge of random office people to deep fake them? Come on...
You're the only commenter using critical analysis, everyone else is just flapping their jaws.
I hate discussing deepfakes. I'm one of the original patent holders of automated actor replacement technology. I developed it for personalized advertising, after having been an actor replacement specialist in a bunch of VFX film you probably saw.
I spent from 2002 to '08 creating a VFX pipeline, with global patent protections, and an ethical guidance that included public education on this fundamental new technology. Long story short, I needed financing, went to VCs and angels and they were perfectly winning to fund a porn company, but not what I'd planned: an ethical rollout of a sensitive and very powerful technology with many legs, few realize even today.
By '13 I was bankrupt, burned out, and one of my tech partners, a global leader in facial recognition hired me. That's a different story. Actor replacement technology is a fundamental capability with applications far more important than fraud and pornography. But our civilization is far far too immature to realize any of them.
> went to VCs and angels and they were perfectly winning to fund a porn company, but not what I'd planned: an ethical rollout of a sensitive and very powerful technology with many legs
Well, yes, that's kind of what the rest of us have come to expect from the industry. Ethical rollout is always going to take a back seat to raking in as much money as possible. I'm slightly surprised they were willing to touch porn though, not for "ethical" concerns but because it's treated as radioactive by payment services.
It was absurd. My refusals began when they would insist on a technology proof that was creating nude celebrities in an image with them as the 2nd person. It was really amazing. Always guys, unable to contain their glee being horny, and insisting, insisting the company make porn. This was every single VC, it met with all of them. Angel investor groups too. It darkened my view of humanity.
> I don't want music recommendations from something that can't appreciate or understand music. Human recommendations will always be better.
I find Spotify's "discover weekly" list to be generally pretty good. Sure, there are some songs I dont like, but there are often 3-4 great songs each week that get added to my regularly playlist.
Its all good an well to say that human recommendations are better, but I'm not paying someone $50 per week to spend 3-4 hours finding me new and good songs. I get something that is maybe 80% as good included, and the reality is that is good enough.
I feel like one of the reason AI is doing well is it doesnt need to be better, it just needs to be "good enough" at a fraction of the price..
>there are often 3-4 great songs each week that get added to my regularly playlist.
I'm earnestly uncertain that a system with near total access to your listening history producing 3-4 great songs per week from the corpus of all human musical endeavour can be considered a "good" effort. Particularly when the 3-4 recommendations are jammed into a 60 minute playlist of otherwise questionable quality.
It’s better than my effort at finding music, and it makes Mondays a little nicer. My goal isn’t to find the best music ever, but to find new and interesting music more easily.
If it's randomly selected from the entire music library of Spotify, then it won't be good. Most of it would come from a long tail of bad or niche stuff.
Honestly, I believe Spotify would offer me more accurate recommendations if they relied solely on AI, without human input. Their current DJ features, although supposedly based on my previous listening habits, often suggest popular songs I don't listen to, tracks supposedly reminiscent of my school days that I've never heard before, and genres that I'm not interested in.
I recently signed up for Qobuz, which costs the same or roughly the same as Spotify. They have a significant amount of recommendations, writing, etc written by actual people. It is of vastly higher quality than anything automatically generated by Spotify. I've only occasionally found something I like through Spotify but have already found many things I like through Qobuz.
But think about it from the end user perspective. Literally the most simple instruction; near fault proof. On an airplane that is thousands of feet from remote IT support (plus "costs").
The instruction to staff; problem with "the Internet"? - press the "Interest Reset" button.
Far better than "router restart", "renew DHCP leases" or "reboot IT"
What's interesting is that the button need not actually reset the Internet right away. It's actually a user signal that "customers are complaining the Internet does not work". The button could initiate a whole series of diagnostics and target a fix.
Honestly, that should be the mindset of IT experts in general. Any reset/reset should fix everything and bring the system to a known functional state before doing any work.
Obviously you don't want to have to restart to fix issues, but having that as a fallback (especially for issues you didn't predict during development) is great UX.
agreed. It’s a great solution for a flight crew that is most likely unable/unwilling to troubleshoot stale DHCP leases or bouncing ifaces. The only disadvantage is that if the Internet reset button doesn’t work for whatever reason, the FAs will mark the entire system INOP for your entire six hour cross country flight that you planned to work on...
So I decided to solve it.
Using the Bosch API - I can tell both when a cycle is complete, and if the door is open. Currently I use their default version, but there is a local hosted option I'll be switching too now the proof of concept works.
So using Home Assistant I have a simple script that detects when a washing machine cycle is complete AND the door has NOT been opened. This implies my washing machine has wet clothes still in it.
So Home Assistant will alert my phone (and my wife only if she is home based upon presence detection) once every 15mins that there are wet clothes waiting in the washing machine.
Very simple - works perfectly.