The Age of Personalized Hardware Is Coming(geastack.com)
geastack.com
The Age of Personalized Hardware Is Coming
https://geastack.com/blog-the-age-of-personalized-hardware-is-coming
89 comments
Witness. I've built three small projects from idea to pilot runs with Ai. Some parts of the process I had some solid experience with, and other parts I was holding the hand of my Ai and hoping he was sober and benevolent. I often had laughing fits of glee when things worked AND I understood them. As good as Ai is at just doing stuff, it's better at explaining and teaching. The back and forth made all the projects better, cheaper, tougher, and ultimately more usable.
For small stuff, sure. For something with more components I don't think it's ever gonna be useful. Routing a pcb is an np-hard problem and, imho, no AI has enough actual thinking capability to make a good job out of it.
The human brain hasn't solved NP-hardness, either. We make it work using heuristics and automation tools, which in turn use heuristics themselves. I see no reason why AI wouldn't be able to take the same approach.
What exactly do you mean by "solved NP-hardness"?
I mean we already have seen this happen with AlphaDev. Not sure why AI would fail at working on NP-hard problems if our current way is to basically guess and keep trying until something clicks (something that is not difficult for an AI system that can just keep running).
An AI that can thrash a Go grandmaster can route a PCB.
There's no reason to think a conventional LLM is the right approach, of course. But the LLM could delegate the task to a program it wrote, given the right prompting and the right feedback loop(s).
There's no reason to think a conventional LLM is the right approach, of course. But the LLM could delegate the task to a program it wrote, given the right prompting and the right feedback loop(s).
What confidence would you have in AI's ability to do 20-40GHz signal routing with good integrity? HDMI + USB4 + USB-C DP AltMode and a bunch of USB4 routing/switching fabric stuff?
Hook it up so it can experiment away by itself with tools, provide a end goal and ask it to iterate until it reached it. What the "AI can do vs not" becomes more about how long you can let it do inference in such loop, with local hardware it gets really cheap (granted you have the hardware already), with remote inference cost is probably the limiter.
> Hook it up so it can experiment away by itself with tools, provide a end goal and ask it to iterate until it reached it
Those "tools" you throw out so cavalierly cost around $50,000 to $300,000 per annual seat license with all the optional packages needed for 40GHz signals (like Cadence Allegro + Sigrity X Speed / SystemSI). Even if AI could do it it'd be much cheaper ($3-8k) to get someone in China to use their license and knowledge to do it.
Those "tools" you throw out so cavalierly cost around $50,000 to $300,000 per annual seat license with all the optional packages needed for 40GHz signals (like Cadence Allegro + Sigrity X Speed / SystemSI). Even if AI could do it it'd be much cheaper ($3-8k) to get someone in China to use their license and knowledge to do it.
But how is cost related to "AI's ability to do X"? Of course if you can't afford the required tools, or inference for that part, remote or local, then you can't do any of that stuff. The same if you don't speak/know the language the specifications/manuals/guides are done in, and you can't afford a translator, but kind of besides the point when we're trying to discuss AI's ability to do something.
For hobby stuff, I'd like to create some boards that involve a bit more than the usual MCU. Things like PCIe or USB4 or HDMI. If AI can't do those things for <$2,000 it doesn't make a difference to my life.
At some point the cost makes it irrelevant. If you drive a Bugatti Chiron, you don't care about the price of petrol. Humans or AI, there's not really a difference if the capital costs vastly outweigh the labor costs.
At some point the cost makes it irrelevant. If you drive a Bugatti Chiron, you don't care about the price of petrol. Humans or AI, there's not really a difference if the capital costs vastly outweigh the labor costs.
What's special about it, other than more constraints, more things to take into account, and having to simulate/test more - so the need for some specialized SW?
They are different constraints than the ones in SW development, but is there some fundamental difference that prevents replacing the human in the loop with AI in the loop?
They are different constraints than the ones in SW development, but is there some fundamental difference that prevents replacing the human in the loop with AI in the loop?
Experienced humans can do basic high speed layout without expensive simulation using rules of thumb that worked for them in the past. It’ll be overconstrained and overbuilt, but it will work.
I’m curious if AI’s know enough to do some useful high frequency / data rate layouts without paying for $50,000+ simulation tools.
I’m curious if AI’s know enough to do some useful high frequency / data rate layouts without paying for $50,000+ simulation tools.
You can certainly make it produce a lot of design rules of thumb. You can make it do the research for each of those based on the literature. And you can then make the machine sort out the rules of thumb by effect and quantify them meaningfully.
Then you can make it process existing good designs and make it propose a general design/routing methodology (and it's own scripts/tools for calculating/optimizing the "thumb" rules that would be used throughout) for initial placement/routing of key components based on all the collected data.
Then you may have enough context and tools to have it make use of them during design.
Then you can make it process existing good designs and make it propose a general design/routing methodology (and it's own scripts/tools for calculating/optimizing the "thumb" rules that would be used throughout) for initial placement/routing of key components based on all the collected data.
Then you may have enough context and tools to have it make use of them during design.
Right. Okay - so you have no idea if it does or not. Thank you.
I actually tried a few of the steps, but whatever.
That's mildly surprising to me, given what I've seen when I ask current models to make an SVG of something.
Would I be on the right track if I guessed there's a DSL for designing PCBs that would help enforce functional correctness?
Would I be on the right track if I guessed there's a DSL for designing PCBs that would help enforce functional correctness?
A lot of human PCB errors could be caught by analyzing the netlist against requirements and knowledge of the datasheet. Schematic and board formats are usually plaintext, so you can generate those directly even. Or kicad + python?
That’s probably enough for an LLM to check if you’ve mis-wired something, missed a part or chose the wrong resistor to set a regulator voltage. Plus good old DRC/ERC. If you pass all of those, there’s a good chance things will work unless your placement is really bad, but you could manually lay out and autoroute a lot of simple boards. Not to belittle the parent but a 4 layer board is actually simpler in some ways because you have a power plane which is one less net to worry about.
For analog work you can even run a SPICE simulation.
That’s probably enough for an LLM to check if you’ve mis-wired something, missed a part or chose the wrong resistor to set a regulator voltage. Plus good old DRC/ERC. If you pass all of those, there’s a good chance things will work unless your placement is really bad, but you could manually lay out and autoroute a lot of simple boards. Not to belittle the parent but a 4 layer board is actually simpler in some ways because you have a power plane which is one less net to worry about.
For analog work you can even run a SPICE simulation.
You got it. It was kicad and python, with the builtin DRC/ERC. It had to iterate quite a lot to get it to pass, but once it did the board was solid, if not elegant. I almost feel like kicad was the real hero.
A human would have done a better job, and maybe used half the PCB... but it got it done and at the end of the day the board is 100% functional, and 100% machine designed.
A human would have done a better job, and maybe used half the PCB... but it got it done and at the end of the day the board is 100% functional, and 100% machine designed.
My attempt at VHDL was a failure but at least it helped me to get a modern build on a Sockit.
But that was a few months ago, getting high hope with fable and seeing killed before I could even try it for that project killed all my motivations.
But that was a few months ago, getting high hope with fable and seeing killed before I could even try it for that project killed all my motivations.
I tried my hand at making a pcb for a project in high power rocketry but I gave up in frustration. Maybe I’ll try again and see if claudecode can help me get over the hump.
Yeah lets just take the human out of the loop in regards to the DIY missile. Sure.
haha not really, so in larger model rockets they go so high that you use two parachutes. The first is deployed at apogee but it's very small so the rocket descends fast and doesn't float (miles in some cases) away, the second is larger and deployed at a pre-set altitude to slow the rocket enough to hit the ground without causing damage to itself or whatever it lands on. To do this you have an onboard altimeter and small blackpowder charges that are used to "pop" out the parachutes.
The altimeters are available OTC but i wanted to use my pi-zero and an IMU with a barometer. What I was designing was an i2c pyro board that could ignite up to 4 e-matches that sets off the blackpowder charges.
I've met the guy that made this altimeter, it's pretty amazing. The hobby is filled with engineers so some of the stuff they come up with is impressive. https://www.featherweightaltimeters.com/blue-raven-altimeter...
The altimeters are available OTC but i wanted to use my pi-zero and an IMU with a barometer. What I was designing was an i2c pyro board that could ignite up to 4 e-matches that sets off the blackpowder charges.
I've met the guy that made this altimeter, it's pretty amazing. The hobby is filled with engineers so some of the stuff they come up with is impressive. https://www.featherweightaltimeters.com/blue-raven-altimeter...
Im surprised by how bad LLMs are with SVGs but somehow are oke-ish with CAD and other weird files.
An SVG is like a cartoon, and throws away a lot of information, a "good SVG" of some subject chosen to be familiar to both animals and humans might look good to one but not the other. The cycling flamingo is a kind of art curation test.
Functional PCB "artwork" is actually not intended for visual consumption, it must meet constraints and advice scattered in documentation, prior "art" in the training corpus, etc.
It doesn't need to be surprising at all
A good static SVG cartoon can be viewed as a multispectral photographic video, with a lot of information thrown away and distorted or oversimplified. Different species will disagree about what information should have been kept vs thrown away. A bee would "complain" it can't see any of the plant's UV markers when presented with plant imagery. Many animals depend on motion or movement to identify predator or prey.
At a certain point culture and personal life experiences start conflicting even within the human subgroup, and we see people argue why this or that flamingo is better than the other.
Functional PCB "artwork" is actually not intended for visual consumption, it must meet constraints and advice scattered in documentation, prior "art" in the training corpus, etc.
It doesn't need to be surprising at all
A good static SVG cartoon can be viewed as a multispectral photographic video, with a lot of information thrown away and distorted or oversimplified. Different species will disagree about what information should have been kept vs thrown away. A bee would "complain" it can't see any of the plant's UV markers when presented with plant imagery. Many animals depend on motion or movement to identify predator or prey.
At a certain point culture and personal life experiences start conflicting even within the human subgroup, and we see people argue why this or that flamingo is better than the other.
For circuits then can be simulated. They have a of constraints that might make the problem space a lot smaller. Maybe there are also a lot of text on what makes a good design.
I also believe most design related to a physical object have documentation justifying the choices.
I also believe most design related to a physical object have documentation justifying the choices.
What do you to get the case? I use build123d with Python and the results are pretty good!
> Then I had it design a case for the new pcb to 3d print.
Can you elaborate on this? What did you use to "design" it?
Im thinking of getting a 3d printer and would love to explore this intersection a bit more.
Can you elaborate on this? What did you use to "design" it?
Im thinking of getting a 3d printer and would love to explore this intersection a bit more.
What harness have you been using for EDA/CAD stuff?
In so far as AI can do hardware reliably you can bet your bottom dollar the big chip fabs have already been doing that. They don’t call it AI though and the models aren’t language based, surprisingly enough /s
The author is missing the _hardest_ part of small hardware: power management.
The reason why AR glasses are not a thing yes is because there isn't a big enough battery to allow them to function for more than a few minutes.
Glasses form factor have space for about .9watt hours now, and ~2 in 4 years time (assuming current trends) Assuming a 14 hour day, that means that your have 140mwhr to spend every hour. A not very bright light on your glasses is about 30mw, a decently bright one is 90mw. Processing imagery on that kind of budget requires custom silicon, and a huge bunch of optimisations.
The reason why AR glasses are not a thing yes is because there isn't a big enough battery to allow them to function for more than a few minutes.
Glasses form factor have space for about .9watt hours now, and ~2 in 4 years time (assuming current trends) Assuming a 14 hour day, that means that your have 140mwhr to spend every hour. A not very bright light on your glasses is about 30mw, a decently bright one is 90mw. Processing imagery on that kind of budget requires custom silicon, and a huge bunch of optimisations.
I think one of the issues is premature optimization/minimzation(?) . All the wearables so far that feature some human interactive display are essentially cramming the mobile concept into different device classes.
Take the AR glasses , they are essentially a computing unit output display , input (joystiicks/sensors) all being crammed into one unit with a powerpack to drive it all. I believe the desktop concept, or in this case 'Walkman' concept would be better fit for wearables .Where one can separate the processing from the display/output and power each separately.Having a cable from your hip pack to the headunit is not so strange/difficult as previous gen of walk/discmans can attest.
However the current ruling desing is wireless everything , super slim devices and all unified with the power supply.
The idea of "pythonizing" resource-constrained environments like embedded seems feasible to me, and even a good idea if executed well (e.g., numpy, torch, jax, etc.), but it does have me cackling imagining the first encounters of modern web devs with performance/power economy constraints
AR glasses coupled to consumer hardware should actually consume less power, since the eye box is relatively small compared to a monitor's "eye box". 99.99% of light emitted from a pixel does not enter the user's pupil for monitors.
Also, I think many consumers wouldn't mind if the rims, frame arms, or frame generally (and optionally a VR mode visor) where covered photovoltaically. if the total area exceeds the area of both pupils (divided by efficiency) the environmental light could power an additive display. (with additive I mean for example pixel-wise LED's so dark doesn't consume power compared to subtractive displays which generate a uniform backlight and then block needless light).
Also, I think many consumers wouldn't mind if the rims, frame arms, or frame generally (and optionally a VR mode visor) where covered photovoltaically. if the total area exceeds the area of both pupils (divided by efficiency) the environmental light could power an additive display. (with additive I mean for example pixel-wise LED's so dark doesn't consume power compared to subtractive displays which generate a uniform backlight and then block needless light).
> AR glasses coupled to consumer hardware should actually consume less power,
Just on that point, if we are talking about a monochrome screen that is just displaying text, that is just an IO device for a mobile phone, I can see the argument but its not really the case.
Screens _eat_ power, waveguides are really not that efficient, so need mini projectors that are overdriven to produce enough light. Current off the shelf for VGA/25 degree is 95mw for the projector only, none of the other hardware. (https://www.jb-display.com/product_des/17.html)
that gives you 4000 nits to the eye. Which is about enough for normal day to day tasks in normal light. (assuming you've cracked waveguides, and thats an active area of research)
But, thats a poor experience, as its not "world locked" as in being able to attach things to real world objects. For that you need a whole new rendering system along with a SLAM stack to know where you are, a machine perception stack to work out what you are looking at. Bear in mind that 90mw for the projector is about 2/3rds of your entire power budget. bear in mind that even Bluetooth LE is quite expensive ranging from 12-40mw for streaming something like 1mb (although through LE thats a tough ask)
Just on that point, if we are talking about a monochrome screen that is just displaying text, that is just an IO device for a mobile phone, I can see the argument but its not really the case.
Screens _eat_ power, waveguides are really not that efficient, so need mini projectors that are overdriven to produce enough light. Current off the shelf for VGA/25 degree is 95mw for the projector only, none of the other hardware. (https://www.jb-display.com/product_des/17.html)
that gives you 4000 nits to the eye. Which is about enough for normal day to day tasks in normal light. (assuming you've cracked waveguides, and thats an active area of research)
But, thats a poor experience, as its not "world locked" as in being able to attach things to real world objects. For that you need a whole new rendering system along with a SLAM stack to know where you are, a machine perception stack to work out what you are looking at. Bear in mind that 90mw for the projector is about 2/3rds of your entire power budget. bear in mind that even Bluetooth LE is quite expensive ranging from 12-40mw for streaming something like 1mb (although through LE thats a tough ask)
just to be clear, I am largely disinterested in standalone glasses for the immediate future for both the precise complaints you aired as well as my lack of need for them to be standalone, I prefer the modularity of display system vs computational platform, which brings its own batteries.
ahh the cyberdeck interaciton model, yeah that actually looks like it could be fun.
I don't know, it's getting better every day.
The other day, I took apart an old handheld vacuum cleaner to scavenge the internals. All there was inside was a small but powerful motor, and a couple of 18650 batteries with a charge controller and a switch.
Two small batteries, running a relatively powerful vacuum, which could do spot cleans of the whole house (corners and crevices and such) without needing a recharge.
The other day, I took apart an old handheld vacuum cleaner to scavenge the internals. All there was inside was a small but powerful motor, and a couple of 18650 batteries with a charge controller and a switch.
Two small batteries, running a relatively powerful vacuum, which could do spot cleans of the whole house (corners and crevices and such) without needing a recharge.
Two 18650s hold about 22Wh of energy, more than two iPhones, and 20x more than what the parent comment is talking about.
While personalization is definitely the trend, I don't think people are going to build code just to personalize. A tiny few of all who bought the device could do that. A few more could flash the device with some open firmware that gives more features and personalization and most will stick with the range of the personalization provided by the vendor.
For the most people, the risks outweigh the desire for tinkering. Personalization will grow right at the vendor offering, not in the hands of customer. People don't even have the time to cook their own recipes. People have their own chores to worry about. I'm talking about bulk of the customer base, not the geeks.
For the most people, the risks outweigh the desire for tinkering. Personalization will grow right at the vendor offering, not in the hands of customer. People don't even have the time to cook their own recipes. People have their own chores to worry about. I'm talking about bulk of the customer base, not the geeks.
I'd include the geeks in that.
I've been programming computers and tinkering with all sorts of hardware for more than 30 years. I first used FreeBSD in.....2001? and Linux not long after that. I've programmed OS code, I've grudgingly written VHDL, I've assembled a sound card for the Apple II I still have running - all this to say that I believe I'm in your tiny few.
And I'm so tired. Tired of having to debug all the things. Tired of having to pay attention to them. Tired of setting them up "just once" and then months later having to reverse engineer my own work because something failed.
So I don't. I leave nearly all my devices stock. I run Windows because I'm sick of debugging device driver issues. And I don't want personalized hardware with any electronics in it (bespoke wooden objects, those I love and make).
I've been programming computers and tinkering with all sorts of hardware for more than 30 years. I first used FreeBSD in.....2001? and Linux not long after that. I've programmed OS code, I've grudgingly written VHDL, I've assembled a sound card for the Apple II I still have running - all this to say that I believe I'm in your tiny few.
And I'm so tired. Tired of having to debug all the things. Tired of having to pay attention to them. Tired of setting them up "just once" and then months later having to reverse engineer my own work because something failed.
So I don't. I leave nearly all my devices stock. I run Windows because I'm sick of debugging device driver issues. And I don't want personalized hardware with any electronics in it (bespoke wooden objects, those I love and make).
I get where you're coming from, though I'm not quite at 30 years yet. I like building stuff, I don't like configuring, debugging or troubleshooting my OS or computer. I just want it to work and get out of the way.
For me this has translated into a slightly different outcome: stock ubuntu with basically just zsh/powerlevel10k, I add component assembly when I buy a new desktop, I'm about to probably replace my home server with a NAS etc.
For me this has translated into a slightly different outcome: stock ubuntu with basically just zsh/powerlevel10k, I add component assembly when I buy a new desktop, I'm about to probably replace my home server with a NAS etc.
Yeah I don't care how exactly my Linux laptop gets two finger scrolling going the right direction, all I know is that I can tell AI what I want and it'll go and change the setting for me.
> I run Windows because I'm sick of debugging device driver issues.
Lost me there. The rest is reasonable but other OSes (MacOS, Linux) also rarely have driver issues. Its not like Windows never needs fixing (these days more than Linux). These days you can buy computers with Linux pre-installed and compatible hardware.
My biggest problem with technology is with very stock stuff. Mobile apps that everyone pushes you to use that clutter up your phone and are often crap. Every time an app I use updates I wonder "what have they broken this time?".
Lost me there. The rest is reasonable but other OSes (MacOS, Linux) also rarely have driver issues. Its not like Windows never needs fixing (these days more than Linux). These days you can buy computers with Linux pre-installed and compatible hardware.
My biggest problem with technology is with very stock stuff. Mobile apps that everyone pushes you to use that clutter up your phone and are often crap. Every time an app I use updates I wonder "what have they broken this time?".
I'm tired of having this debate over and over. I still use Linux devices regularly at work and have a T420S and various RPi-like devices that run it, so it's not like I'm going off memories of 2010. I owned a Framework Laptop (designed to run Linux) and got rid of it in no small part because the battery life was so much worse on Linux than it was in Windows. I've worked at Google who (circa 2019) had an incredible team of IT specialists and specifically selected hardware and still struggled to get Bluetooth on Linux to work reliably for everyone.
I don't want to have to carefully pick hardware for compatibility. I want first class support for pen and touch. I value thin, light, and long battery over the ability to run a custom bootloader.
Here's a thread where I ranted with specific examples 3 months ago: https://news.ycombinator.com/item?id=47463982
I don't want to have to carefully pick hardware for compatibility. I want first class support for pen and touch. I value thin, light, and long battery over the ability to run a custom bootloader.
Here's a thread where I ranted with specific examples 3 months ago: https://news.ycombinator.com/item?id=47463982
I am tired of arguing about this as I do not not have these problems and have not in many years over multiple machines.
Looking at your specific examples they are not a like for like comparison. Two are Ubuntu on Raspberry Pi issues (so on a device designed for experimentation, and not even running the usual distro). One is a bug in the GUI for True NAS. One is unhappiness with the default installed command line packages for Debian on WSL.
When I used Windows I had just as many irritants. It was very noticeable when I was using Linux at home and Windows at work (and is one reason I switched to Linux). it is definitely not true that Windows always works flawlessly for everyone in all circumstances. If you count every irritation you find with every Linux distro on a wide range of hardware it will add up to more than Windows on typical hardware. Have your tried Windows on your Pi? Have you replaced your TrueNas server with a Windows equivalent?
Looking at your specific examples they are not a like for like comparison. Two are Ubuntu on Raspberry Pi issues (so on a device designed for experimentation, and not even running the usual distro). One is a bug in the GUI for True NAS. One is unhappiness with the default installed command line packages for Debian on WSL.
When I used Windows I had just as many irritants. It was very noticeable when I was using Linux at home and Windows at work (and is one reason I switched to Linux). it is definitely not true that Windows always works flawlessly for everyone in all circumstances. If you count every irritation you find with every Linux distro on a wide range of hardware it will add up to more than Windows on typical hardware. Have your tried Windows on your Pi? Have you replaced your TrueNas server with a Windows equivalent?
> Personalization will grow right at the vendor offering
I think so. There will be a short period, like now, where many will attempt to build their own products and the five good ideas out of five thousand will be incorporated into products built by those who know what they’re doing.
An absolutely technophobic friend asked me what agentic AI was today. I think Charlie Munger said “When even your barber is talking about it, sell and run”
I think so. There will be a short period, like now, where many will attempt to build their own products and the five good ideas out of five thousand will be incorporated into products built by those who know what they’re doing.
An absolutely technophobic friend asked me what agentic AI was today. I think Charlie Munger said “When even your barber is talking about it, sell and run”
I'd bet Charlie Munger and his barber talk about cars and don't run. When we're talking random stock tips that's one thing but if it's possibly a foundational society changing technology like plastic, maybe your barber talking about it isn't the "get out now" signal that neat little story purports to be.
Nobody is questioning the longevity of AI. The internet itself wasn’t affected by pets.com but a lot of retail investors got burned. It’s going to happen again, I think perhaps in a different way.
For example many are treating vibe coding like slot machines and the app monetisation gamble is unlikely to pay off.
For example many are treating vibe coding like slot machines and the app monetisation gamble is unlikely to pay off.
The keyboard enthusiasts got to "personalized hardware" pretty early, so there might be some lessons there.
There's a whole continuum from "Buy an off the shelf unit" to "here's a barebones case you can slide ready made switches and caps into" to "mix and match custom PCBs and cases" to designing your own PCBs and/or cases. There are pretty clear pipelines and even some levels of tooling for "draw up the layout you want and get a bunch of files you can send to production houses"
Takeaway 1: A lot of this is grounded in economic realities. I did the full bespoke route (custom PCB, custom 3-D printed case) and figured it probably cost me about $500 all inclusive to get what I wanted, and that's honestly a lot of money for a keyboard.
AI won't solve any of the economic problems. They can't fix "the minimum PCB order is N units, so now you have a drawer full of spares you paid for". They can't make the expensive part you needed cheaper, especially if you're an individual buying quantity of 1/5/10 instead of an OEM buying reels and containers-full. They can't change the fact that a case for a large widget will be expensive to mill/3-D print/mould/etc.
Takeaway 2: Customers may have surprisingly limited imagination for bespoke gear. There are galleries (and even coffee-table books) full of exotic keyboards. But Micro Center is full of $50 interchangeable "tenkeyless with RGB lighting" boards; throwing on a random set of "custom" keycaps, and that's enough for a large part of the audience.
Will these customers want or benefit from more tools, or will it just give them rope to hang themselves on and give them an excuse to bail out of the purchase entirely? Even if you can provide them a gallery of vetted turnkey choices, there might be more choice paralysis than actual benefit.
Takeaway 3: Hardware is forever (relative to software). You have a lot of small firms and group-buy products that disappeared and now the owners can't get an exact replacement or repairs. Conversely, Unicomp can gut and rebuild a 1986 Model M with new innards in large part because they've been selling the same basic design since a 386DX/16 cost as much as a Toyota Tercel.
If your AI spawns a galaxy of 1-of-1 bespoke products, who services and supports them? That seems like it's only going to appeal to the enthusiast-hacker type who can keep them alive themselves, who is least likely to need AI help designing them. Design for disposability isn't a great look for anything but incredibly low-cost, limited-usecase items.
There's a whole continuum from "Buy an off the shelf unit" to "here's a barebones case you can slide ready made switches and caps into" to "mix and match custom PCBs and cases" to designing your own PCBs and/or cases. There are pretty clear pipelines and even some levels of tooling for "draw up the layout you want and get a bunch of files you can send to production houses"
Takeaway 1: A lot of this is grounded in economic realities. I did the full bespoke route (custom PCB, custom 3-D printed case) and figured it probably cost me about $500 all inclusive to get what I wanted, and that's honestly a lot of money for a keyboard.
AI won't solve any of the economic problems. They can't fix "the minimum PCB order is N units, so now you have a drawer full of spares you paid for". They can't make the expensive part you needed cheaper, especially if you're an individual buying quantity of 1/5/10 instead of an OEM buying reels and containers-full. They can't change the fact that a case for a large widget will be expensive to mill/3-D print/mould/etc.
Takeaway 2: Customers may have surprisingly limited imagination for bespoke gear. There are galleries (and even coffee-table books) full of exotic keyboards. But Micro Center is full of $50 interchangeable "tenkeyless with RGB lighting" boards; throwing on a random set of "custom" keycaps, and that's enough for a large part of the audience.
Will these customers want or benefit from more tools, or will it just give them rope to hang themselves on and give them an excuse to bail out of the purchase entirely? Even if you can provide them a gallery of vetted turnkey choices, there might be more choice paralysis than actual benefit.
Takeaway 3: Hardware is forever (relative to software). You have a lot of small firms and group-buy products that disappeared and now the owners can't get an exact replacement or repairs. Conversely, Unicomp can gut and rebuild a 1986 Model M with new innards in large part because they've been selling the same basic design since a 386DX/16 cost as much as a Toyota Tercel.
If your AI spawns a galaxy of 1-of-1 bespoke products, who services and supports them? That seems like it's only going to appeal to the enthusiast-hacker type who can keep them alive themselves, who is least likely to need AI help designing them. Design for disposability isn't a great look for anything but incredibly low-cost, limited-usecase items.
Mmmmmmmmmm… sort of.
If you are keeping inside the LEGO level of complexity, almost. But just as with code, if the project is complex you need a real engineer herding the cats. That said, it certainly can extend the reach of an effort, and is a huge help doing board reviews and data sheet analysis, etc. but the real engineering decisions are very very hit and miss, just like with code.
If you are keeping inside the LEGO level of complexity, almost. But just as with code, if the project is complex you need a real engineer herding the cats. That said, it certainly can extend the reach of an effort, and is a huge help doing board reviews and data sheet analysis, etc. but the real engineering decisions are very very hit and miss, just like with code.
This reminds me of the other article posted on here, https://isene.org/2026/05/Audience-of-One.html Everyday more bespoke dotfiles are coming online for an audience of one. I am not surprised that this is coming to hardware as well.
AI + hardware has really helped my wife and I get more sleep.
I had an esp32-box-3 lying around from a lapsed "voice agent" project from a year or two ago. Had a baby. Baby moved to another room, sleep trained. Baby either: 1. wakes up a few times a night, babbles for a bit, goes back to sleep OR 2. baby wakes up and fusses for N (=10) minutes, at which point parents need to go in and settle (that's the sleep training routine we use).
In either case, we do NOT want to wake up every time the baby does. Baby can go back to sleep easily, we adults have a harder time. A few rounds with Claude and the esp32 is now our new baby monitor. It tracks cry/fuss duration and publishes an audio stream (via a web UI or direct with, say, VLC). The audio only comes through AFTER N minutes of fussing have elapsed. It also posts notifications (to ntfy) after 30s and N minutes. My log says baby often wakes up 1-2 times a night and resettles almost immediately. We only wake up if the audio comes through, after N (10) minutes.
Also during the day it's really handy to be notified when baby has woken up from her nap. Let's us be out of the house, or in a distant room, and still keep track of what's going on.
It's fun to keep improving and adding features to this. Never would have had the time/energy to get this done without a coding agent. I ordered a set of 10 more of the esp32-box-3s to give them out to my friends (well, some are for other projects... so much potential).
(EDIT: Yes, I know this isn't AI designing hardware, but even writing code for embedded off the shelf stuff feels like a huge new potential.)
I had an esp32-box-3 lying around from a lapsed "voice agent" project from a year or two ago. Had a baby. Baby moved to another room, sleep trained. Baby either: 1. wakes up a few times a night, babbles for a bit, goes back to sleep OR 2. baby wakes up and fusses for N (=10) minutes, at which point parents need to go in and settle (that's the sleep training routine we use).
In either case, we do NOT want to wake up every time the baby does. Baby can go back to sleep easily, we adults have a harder time. A few rounds with Claude and the esp32 is now our new baby monitor. It tracks cry/fuss duration and publishes an audio stream (via a web UI or direct with, say, VLC). The audio only comes through AFTER N minutes of fussing have elapsed. It also posts notifications (to ntfy) after 30s and N minutes. My log says baby often wakes up 1-2 times a night and resettles almost immediately. We only wake up if the audio comes through, after N (10) minutes.
Also during the day it's really handy to be notified when baby has woken up from her nap. Let's us be out of the house, or in a distant room, and still keep track of what's going on.
It's fun to keep improving and adding features to this. Never would have had the time/energy to get this done without a coding agent. I ordered a set of 10 more of the esp32-box-3s to give them out to my friends (well, some are for other projects... so much potential).
(EDIT: Yes, I know this isn't AI designing hardware, but even writing code for embedded off the shelf stuff feels like a huge new potential.)
This is one of the killer use cases of AI, build personalized stuff for your life, that no company does. It is kind of generalized „intelligent stuff“ that one can use. Like minecraft your life
This actually feels like something that could be a popular app on a baby monitor.
Baby monitors fundamentally need to be safe and allow parents peace of mind. Knowing they can constantly hear the baby and communicate verbally to soothe, as if you were in the room with them, allows this. It already means you only have to respond and/or get up to tend if they are still unsettled after a period of time. Monitors also have volume control so you can stay alert without being constantly disturbed.
The traditional baby monitor system had three states (ON and functional, ON and non-functional, OFF). It provides a constantly available and instantaneous test for "functional" - as long as you can hear sound from the other unit, it's almost guaranteed to currently be in a safe condition. Monitoring is constant, human-first, and all human. The system is relatively fail-safe.
Your design replaces this with a multi-state system and algorithms (adding ON in listening mode, ON in delay mode with output off, ON and partially-functional etc.). It removes any reliable method to prove an obviously "functional" state at any given time. Monitoring is non-constant (for the human), machine-first, and human as last resort (like corporate customer services). The system is not fail-safe despite being machine-first and at higher risk of error or malfunction due to complexity.
The risks include a few failed notifications or incorrect delay timing leading to early developmental trauma, such as fear of abandonment. It doesn't take much. We are still learning to deal with this human-human, before adding invisible unknowns.
If nothing else, what are babies likely to learn in a pre-verbal state where the days are spent sensing and observing their environment to develop the brain? It'll probably be how to game the baby monitor algorithm.
> Also during the day it's really handy to be notified when baby has woken up from her nap. Let's us be out of the house, or in a distant room, and still keep track of what's going on.
If you are going out of earshot of the baby by doing this, you are fully relying on the technology being functional. That would seem unnecessarily risky, and not nice for the baby (they can sense this stuff). The odds of a catastrophe are low in a singular environment, but still enough to worry. Almost inevitably it would end badly if this were scaled to a mass consumer product.
The idea is interesting from an engineering view, but from a human one it feels dystopian to insert a machine between parent and baby to this extent. It removes/replaces a layer of human-human connection. Where does engineering the natural human experience out of life end? Automated feeding? AI nannies and teachers? Then onto AI therapist?
My suggestion would be to do a lighter version of the features in a system which focuses on safety (always being able to monitor no matter what, for example on the ESP32, you could have the second core independent and direct-output the feed if the first core hangs or crashes). Feature-wise, rather than not alerting for 10 minutes, you could apply a DSP algorithm which reduces harsh frequencies for this period, whilst slowly increasing the output volume of the monitor.
I like the idea of alerts because it expands the base features without risk, and crucially makes the babies life better, as alerts could be sooner for important things that do need attention quickly. As an experiment, you could add an SD card and record snippets of cries which ended up in alerts, tag them with an ordered list of what you did to soothe or what was wrong, and see if AI can make anything of the data. Maybe certain cries can trigger alerts sooner, or cries over an extended period may indicate fever.
The traditional baby monitor system had three states (ON and functional, ON and non-functional, OFF). It provides a constantly available and instantaneous test for "functional" - as long as you can hear sound from the other unit, it's almost guaranteed to currently be in a safe condition. Monitoring is constant, human-first, and all human. The system is relatively fail-safe.
Your design replaces this with a multi-state system and algorithms (adding ON in listening mode, ON in delay mode with output off, ON and partially-functional etc.). It removes any reliable method to prove an obviously "functional" state at any given time. Monitoring is non-constant (for the human), machine-first, and human as last resort (like corporate customer services). The system is not fail-safe despite being machine-first and at higher risk of error or malfunction due to complexity.
The risks include a few failed notifications or incorrect delay timing leading to early developmental trauma, such as fear of abandonment. It doesn't take much. We are still learning to deal with this human-human, before adding invisible unknowns.
If nothing else, what are babies likely to learn in a pre-verbal state where the days are spent sensing and observing their environment to develop the brain? It'll probably be how to game the baby monitor algorithm.
> Also during the day it's really handy to be notified when baby has woken up from her nap. Let's us be out of the house, or in a distant room, and still keep track of what's going on.
If you are going out of earshot of the baby by doing this, you are fully relying on the technology being functional. That would seem unnecessarily risky, and not nice for the baby (they can sense this stuff). The odds of a catastrophe are low in a singular environment, but still enough to worry. Almost inevitably it would end badly if this were scaled to a mass consumer product.
The idea is interesting from an engineering view, but from a human one it feels dystopian to insert a machine between parent and baby to this extent. It removes/replaces a layer of human-human connection. Where does engineering the natural human experience out of life end? Automated feeding? AI nannies and teachers? Then onto AI therapist?
My suggestion would be to do a lighter version of the features in a system which focuses on safety (always being able to monitor no matter what, for example on the ESP32, you could have the second core independent and direct-output the feed if the first core hangs or crashes). Feature-wise, rather than not alerting for 10 minutes, you could apply a DSP algorithm which reduces harsh frequencies for this period, whilst slowly increasing the output volume of the monitor.
I like the idea of alerts because it expands the base features without risk, and crucially makes the babies life better, as alerts could be sooner for important things that do need attention quickly. As an experiment, you could add an SD card and record snippets of cries which ended up in alerts, tag them with an ordered list of what you did to soothe or what was wrong, and see if AI can make anything of the data. Maybe certain cries can trigger alerts sooner, or cries over an extended period may indicate fever.
Thank you for engaging with this thoughtfully. You've highlighted the HUGE difference between "bespoke just for me/us" and "product that works for everyone". I've been struggling with this myself, since its value to my family is such that I immediately think "how can I share/sell this to everyone?"
In my case, I have a separate heartbeat check process (running on another machine) which ensures that the monitor is correctly working (no silent fails). I also wired up a SECOND esp32-box-3 as a receiver, and it will complain loudly if the transmitter one isn't sending (much like most normal baby monitors do). The monitor (web ui) has an option of directly listening to the raw non-gated audio, which I use every now and then to confirm what's going on. I also record the streams sometimes for algorithmic improvement.
My aim is not to replace human connection with machines, it's to allow the parents a bit more sanity so as to be better carers for our children. So far we've gained in sleep, piece of mind, and reduced stress. The last one is important - listening to your baby yell for 10 minutes, even if you know they're just annoyed because they don't want to sleep, can be really draining (especially for mom). The notifications allow us to keep tabs on things without the direct audio line to the limbic system.
In my case, I have a separate heartbeat check process (running on another machine) which ensures that the monitor is correctly working (no silent fails). I also wired up a SECOND esp32-box-3 as a receiver, and it will complain loudly if the transmitter one isn't sending (much like most normal baby monitors do). The monitor (web ui) has an option of directly listening to the raw non-gated audio, which I use every now and then to confirm what's going on. I also record the streams sometimes for algorithmic improvement.
My aim is not to replace human connection with machines, it's to allow the parents a bit more sanity so as to be better carers for our children. So far we've gained in sleep, piece of mind, and reduced stress. The last one is important - listening to your baby yell for 10 minutes, even if you know they're just annoyed because they don't want to sleep, can be really draining (especially for mom). The notifications allow us to keep tabs on things without the direct audio line to the limbic system.
> Here's the catch. To put a decent interface on a seven-dollar board today, you still need C++, a board SDK... None of that looks anything like building for the web, so the way most people build interfaces stops at the edge of the device.
So they put a web browser in the device?
So they put a web browser in the device?
I’ve had Claude build UIs from scratch in rust. No framework. I prompted it to make its own ImGUI style Ui system and a few minutes later I had texts, icons, buttons, sliders, scrolling lists, etc..
All my experience tells me it can do it with or without a GPU meaning if you don’t have one it can easily write a software render for a UI
All my experience tells me it can do it with or without a GPU meaning if you don’t have one it can easily write a software render for a UI
Here is the author of Gea Stack. There is no browser in the device, none of these devices could run one with 512KB of RAM (+2-8MB of PSRAM). We instead transpile TypeScript and CSS into native code, so the UI you build with web technologies look and behave identical on a microcontroller.
And it's probably going the same way all the personalized software we've been seeing: used as a spam "I built this" post on reddit or hacker news and to never be touched again.
With hardware you get extra safety risks of fires and shocks, so let's see
With hardware you get extra safety risks of fires and shocks, so let's see
I won't submit to data sniffing from corporation,
so "personalized" hardware won't make it into this
area here.
By the way, this is not new either. In the 1990s Microsoft tried to convince people how great it is if their fridge gives data to MS so MS can order needed food etc..
Fast forward some decades - many don't want to yield any more data to private entities, no matter which alleged "benefit" this would bring.
By the way, this is not new either. In the 1990s Microsoft tried to convince people how great it is if their fridge gives data to MS so MS can order needed food etc..
Fast forward some decades - many don't want to yield any more data to private entities, no matter which alleged "benefit" this would bring.
I feel like the article ignores the elephant in the room — production costs.
Producing PCBs (say, 5 or 10 units) is pretty expensive, and components are also costly on such a small scale. Combining these two requires additional money or time. Beyond that, you need to consider that you probably won't get it right the first time, and every attempt multiplies the cost.
It'll be a deal-breaker for many people — the risk is just too high.
Producing PCBs (say, 5 or 10 units) is pretty expensive, and components are also costly on such a small scale. Combining these two requires additional money or time. Beyond that, you need to consider that you probably won't get it right the first time, and every attempt multiplies the cost.
It'll be a deal-breaker for many people — the risk is just too high.
Pcbs are basically rolling off a laser jet printer in shenzhen and they’ll do pnp for a few dollars more. Pcb cheap.
Have engineer sit there to design pcb is expensive
Have engineer sit there to design pcb is expensive
Last time I tried JLCPCB, it was way too expensive on a smaller scale. Simple PCB with a cheap Nordic nRF, a few other basic components, 10cm x 5cm: ~17 euros/unit (including shipping) at 5 units which was the minimum amount to order.
Do you have a different experience?
Do you have a different experience?
Yes, raw PCBs that size are usually closer to 1EUR/unit. DHL shipping on these slightly larger than average boards would be 3EUR/unit.
Are you talking about fully assembled (pcb + components + solder mask + pnp + soldering + test) at 17EUR/unit? If so, that's incredible cheap for a finished product delivered to your door. If not, you're doing something wrong.
Are you talking about fully assembled (pcb + components + solder mask + pnp + soldering + test) at 17EUR/unit? If so, that's incredible cheap for a finished product delivered to your door. If not, you're doing something wrong.
With hardware/computers becoming locked, thin clients and unaffordable we're already forced to customize our devices cyber punk style.
You're lucky if you're in a region where these open-hardware companies sell their wares, even though many of them will go under in the current market.
You're lucky if you're in a region where these open-hardware companies sell their wares, even though many of them will go under in the current market.
I recently grabbed a cheap e-reader and installed a fork of the firmware so I could use it as a dumb terminal for an agent. Hackable firmware lets you totally change a product
Instead of staring down at their phones in social situations people can pretend to pay attention wile looking at a screen. Can't wait for zombie stairs.
Ehhh, maybe? I haven't heard many people wishing for devices that were within the realm of practicality (I.E. not a flying car). If someone has an actual good idea, then the cost of the components and tools and rework is the next major hurdle (soldering by hand is where cool ideas go to die). Meanwhile a commercial off-the-shelf device probably works good enough.
> People build their own applications now
All the people that I know in tech and not in tech, don't do that, even if presented with the option they're too busy to also get into this endeavour that still requires a lot of expertise. I stopped reading there
All the people that I know in tech and not in tech, don't do that, even if presented with the option they're too busy to also get into this endeavour that still requires a lot of expertise. I stopped reading there
"So the interesting question isn't whether personal hardware is coming. It's who gets to write the software that runs on it."
It's not that I don't like your point of view. It's that I can't stand AI slop.
It's not that I don't like your point of view. It's that I can't stand AI slop.
I love the slider UI element to switch between different examples on their landing page (https://geastack.com/)... not sure if satire or real.
well my phone can do almost everything
anything but be owned by you
No, by reasonable definitions that's doable too. My phone runs an OS I chose, that I have admin access on, that runs any app I tell it to run. And, y'know, it's my property that I bought with my own money, but that's probably aside your argument.
I do agree that hardware gap starts to shrink, similarly to what Software gap once used to be. It is much easier to avoid spending time loading the right drivers and looking up what error message on that tiny device you are working on means. Especially if AI could run in terminal itself.
I tried building a health device few years ago and got completely lost in how to setup camera and touch display with a raspberry PI. Would imagine, with AI running in a command line, it would be much easier.
I tried building a health device few years ago and got completely lost in how to setup camera and touch display with a raspberry PI. Would imagine, with AI running in a command line, it would be much easier.
I had Claude design an entire 4 layer rp2040 based PCB from scratch and PCBWay build it. It worked on the first go, other than some silkscreen overlapped, which doesn't hurt anything. That was before Fable.
Then I had it design a case for the new pcb to 3d print. Also worked the first go, but with minor cosmetic issues.
People have yet to even BEGIN to appreciate what these things can do with the right harness.