Can AI design circuit boards yet?

Posted by iopapa 4 days ago

Comments

Comment by SequoiaHope 4 days ago

I have 15+ years of PCB design experience. Mostly hobby stuff but a fair amount of processional work. Kilowatt range brushless motor controllers, basic RF stuff, lots of microcontroller stuff.

I had Fable design an LED earring. Rechargeable coin cell, RP2350 cpu, IMU, 45 addressable LEDs. It made two mistakes - missed the through holes on the coin cell holder footprint and made the center pad too small. I was able to have JLC swap the through hole battery holder for a surface mount one, and I put a little solder on the small center pad to make it stick up above the mask. They work great! It took 6 days of Fable usage, so about $50 on my Max plan. Very cheap for hardware dev.

I was sufficiently impressed that I’ve been going over old circuit board designs. Some half finished, some completed but in need of a next rev, and I’m getting so much done.

To see it hit the mainstream like the OpenAI announcement, I think big things are coming for this world and by and large they are not ready for it.

For my part, I have always loved PCB design and layout but I simply can’t keep up with the amount of labor required to build what I want, so I welcome this change.

I have also begun exploring more advanced algorithms for PCB manipulation. I have a fairly dense board that needs a few more small chips added. I have an algorithm now that can kinda shuffle and jostle things around so you take up all the spare microns of space across a region of the board and make openings to squeeze a little more in there. It’s pretty cool to see the visualizations as I have it generate movies of the component drift. I foresee much more powerful tools like this in the future.

One tip: have it make a project web page with a chronological list of big changes and detailed visualizations for everything that happens. I can actually prompt all of this on my phone while I am out and about, and view the results on a Tailscale served local page. I’ve always wanted to be able to do PCB design when away from home and now I can!

Comment by zorm 3 days ago

As someone without PCB design experience but with dreams of physical things that could be made with PCBs this sounds amazing. What software are you using for the PCB design? Or just entirely letting Fable do whatever and checking the work with the browser page you mention?

Comment by SequoiaHope 3 days ago

The designs are in Kicad format, and the Kicad APIs are useful for Claude to work with. I would say that right now this involves a lot of knowledge about PCB design to do well, or for more detailed designs. However you might research how to design board for fabrication at JLCPCB and try something simple. Eg make a board with only LEDs in some pretty pattern and make it so you solder a raspberry pi Pico 2 to the board to drive it. Then you don’t have to worry about microcontroller work, just getting the LEDs placed. That is a little easier to get right. Just check the orientation of the LED footprint in the data sheet and make sure the four pins on the LED footprint match - and that there is a silkscreen dot on pin 1 of the LED. You can feed an agent this comment and it can check, though it’s good to check with your own eyes. You might ask it to screenshot the data sheet and the board footprint and display them side by side on the project web page for your review.

Put the LEDs in a cool pattern, slap a Pico on there, have your LLM program it, then dangle it off your backpack with a USB battery pack. Probably adding a motion sensor is easy enough.

Ask the LLM to make sure the board follows JLCPCB’s design rules, lists the LCSC part number for each part, verifies the parts are in stock, double checks every footprint, and makes sure the board passes DRC. Make sure it creates a schematic that is linked to the board design, and that the schematic is properly arranged in to logical blocks with clear connections the way a person would make a schematic - not a big array of parts with global labels for everything. Once it’s done, ask it to clean up the schematic and make it better. Repeat for the board layout. Ask it to review and look for issues multiple times. It will find them. Finally, take some time to doodle your own silkscreen art on there. Have fun! Note the Pico 2 is USB micro. For USB C, sparkfun or waveshare sell similar boards.

Also you can ask it to teach you! Ask a million questions and have it give you multimedia explainers.

Comment by Always42 3 days ago

I have both Altium and Kicad experience. Can you explain further how you are setting up the AI to interface with KiCad? Is your AI having keyboard/mouse access or are you doing it only through API? Not familiar with Claude interface.

Comment by moggi 3 days ago

I have had good success with Claude Code combined with https://github.com/mixelpixx/Konnect. I still need to do routing manually but it is quite useful for reviews of the design and even of the final PCB.

Comment by SequoiaHope 3 days ago

I have it convert Altium files to Kicad and then it uses the Kicad api to control things. If you give it encouragement it should be able to convert back to Altium by parsing the format and building a tool. Supposedly I have this built though I have not validated it. You can also ask it how it would do this and it will have ideas.

Comment by stavros 3 days ago

I'd like the AI to do the schematic for me, as I always worry I've missed some decoupling capacitor or something. I can do the routing myself (I like routing). It would also be good if the AI could fill in all the LCSC identifiers for assembly. Is that doable nowadays?

Comment by SequoiaHope 2 days ago

Yes! They’re good at reading data sheets and finding things to add, locating LCSC parts, and making sure they’re connected in the schematic. They don’t always make a pretty schematic but they can be convinced to make the look a bit better than their default attempts will produce.

Comment by stavros 1 day ago

Excellent, thank you! I'm definitely adding an LLM review/assembly step then!

Comment by varispeed 3 days ago

Will it improve already routed board? I have one done ages ago. Works fine, but RAM can only do roughly a half bandwidth because of poorly laid traces. Could I tell AI to reroute them according to best practices?

Comment by SequoiaHope 3 days ago

Ya it can

Comment by varispeed 2 days ago

sweet

Comment by varenc 3 days ago

> I think big things are coming for this world and by and large they are not ready for it.

With regard to PCB design, I'm curious if can you add more color to how the world isn't ready for this in particular? I don't think I quite grasp all the ramifications of AI becoming PCB design experts.

Comment by tomhow 2 days ago

I can speak from experience here (I’ve been around small-scale electronics design/manufacturing for decades via my father’s business and was hands-on with it on an ag-IoT project for about 10 years).

The limiting factor in electronics/PCB development and manufacturing isn’t engineering expertise (there are plenty of sufficiently competent electronics engineers in the world, especially in Asia), nor parts/materials or assembly costs. It’s the endlessly laborious grunt work of testing a design and making minor improvements until it is finally ready for volume production. That process can easily take a year or two even for a fairly simple design. Even a really skilled/experienced engineer can make a simple mistake in a schematic or PCB layout, which only gets picked up weeks later when prototypes are assembled and tested, or even later when a batch has been made and shipped to customers. So, each design-assemble-test cycle takes weeks/months, and each testing/diagnosing/repairing session takes hours or days. And it mostly has to be undertaken by degree-qualified engineers or experienced technicians.

And that’s mostly before you’ve written all the firmware, which is incredibly laborious in its own right.

The outcome is that only very large companies with big sales volumes can afford to do it (or niche players able to sell at huge margins, most commonly in defense or medical industries).

Outside of those sectors, it’s now just not really viable to be a small-medium business selling moderate volumes of units anymore, the way it was in the 80s-90s. I’ve seen it up close in agricultural IoT. There are no big/dominant players (customer needs are too heterogeneous and distribution is so hard). There are many small vendors, but they find it hard to make products that work well and that customers love, especially at a price point most growers will go for. So, no vendors are big commercial winners and not many customers are very happy - even though there’s a great need for technology in this sector. And it’s really not because the vendors and their engineers are incompetent; it’s just really hard.

With AI to take care of so much of the grunt work involved with circuit/schematic design, PCB layout, pre-manufacturing verification, testing, and firmware dev, it’s realistic to me that you could cut down on R&D costs by 90%, make it much faster to get products onto the market, and make it possible for so many more high-quality niche products to be commercially viable.

Comment by bsder 2 days ago

Unfortunately, my experience has been that it's the endless cycling with customer requirements that takes all the time.

The hardware is generally manageable quickly for me. Even th firmware isn't terrible.

And then you deploy it.

And the customer realizes it needs to do Q and that they asked it to do X and it doesn't need to do that and we need to take another $0.50 out of the BOM and ...

And then 4 years pass and you still don't have any real revenue.

There is a reason why so many places demand NRE of $1 million minimum and THEN they'll talk about designing your project.

The big gap isn't scaling. It's the scaling PATH. You need to have a viable and profitable path all along AS you scale from tens to thousands to millions. And those projects are VERY rare.

See: BeatBuddy for interesting post mortem and note that AI wouldn't really have helped them much at all.

Comment by rjh29 2 days ago

I don't understand your argument. PCB design is not that hard you can learn how to do it in a month or so. JLCPCB can print (or even assemble) PCBs for next to nothing. Irrespective of AI it seems easier than ever to produce prototypes right now.

Comment by tomhow 2 days ago

Yes, you can learn PCB design “in a month”, design your board and get a batch of prototypes made up by JLCPCB. All of that takes 2-3 months absolute minimum. Then you have to test the boards. You need a lot of costly equipment and expertise to test the board and fix any problems with it. That can take days to weeks of work. Then ordering your next batch of prototypes carries a 2-4 week turnaround time including shipping. Repeat this for any design error or just any improvement you want to make once you’ve tried working with it for a bit.

Then you have to write the firmware. That’s weeks or months of work. (And much of the embedded firmware code I’ve seen come out of small business engineering departments or consulting shops is terrible, because they don’t have the time/budget or organizational structure that delivers clean, readable, well-structured, well abstracted C code.)

All this is before you get an MVP into the hands of your customers. Then you start getting customer feedback and have to start making more improvements.

“Never been easier” it may well be to get a hobby project prototypes. But for a product with any amount of serious functionality, it’s at least months and likely 1-2 years to get anything into volume production.

But the potential for saving huge amounts of time particularly on design/layout, pre-manufacturing error detection, and production of well-structured and debugged firmware code are huge.

Comment by rjh29 2 days ago

You lost me when you referred to a multimeter as "costly equipment"

Comment by tomhow 2 days ago

I don’t know why you need to be snarky. I’ve been living this for over a decade with engineers with decades of experience who have done some of the most advanced electronics work imaginable (satellites, military drones).

To test and rework a surface-mount PCB, you may need:

- logic analyzer or oscilloscope;

- microscope;

- high precision soldering equipment/ability;

- solder paste and oven;

- cutting equipment to cut through 4-6 PCB layers and connect tracks with hair-width wire;

- Years/decades of education and experience to understand what to look for and what to do when you find a problem.

Comment by rjh29 2 days ago

If you made the PCB then you should be well placed to fix issues with it. You don't need years of experience , you have the schematic. A microscope (if you need it) is 40 USD. A logic analyser is 10 USD. Solder paste and hot air is only for specific parts and that is 30 usd.

I am assuming 2 layer - 6 layer PCBs are very specialised (e.g. drones)

The reason I'm being snarky is because it feels like you don't know what you're talking about. I know you're a mod so ban me or we but that's the impression I got.

Comment by foldr 2 days ago

As tomhow points out, you are likely going to need more than just a multimeter. However, the hard part isn't whipping up a board that's 90% of the way there; it's the last 10%. I'm currently on revision ~11 of a hobby project which is far less complex than anything tomhow is talking about (4 layer PCB, 8-bit microcontroller, one analogue sensor circuit). Even the first revision of the board 'worked' for the most part, so I'm clearly not totally incompetent. It's just that the kind of polish that takes you a day or two of iteration if you're designing a web frontend can take months (or years, if you're working in your spare time) when you're constrained by manufacturing lead time and the need to test a physical object. There's a huge gap between "Yay! LED flashes!" and a marketable niche consumer product. Even just successfully putting the bloody thing in a box can require a lot of trial and error.

The firmware is not to be underestimated either. In my case that's about 6000 lines of C. That doesn't sound like much, but writing, testing and updating C code targeting an 8-bit microcontroller is a whole different ball game to writing generic application code. The time investment per line is vastly higher. (In fairness, choosing an EFM8 instead of an 32-bit ARM part was probably a mistake on my part.)

Comment by PennRobotics 1 day ago

I recently layed out and ordered a single-sided two-layer board with only a minimum config STM32 + crystal, two USB ports with ESD protection, and two LEDs (one is always-on). I went for commodity SMD parts and it still costs 20 USD per unit to produce and ship from China. Even that PCB doesn't work right now.

[Sorry, dear reader. This post turned into a rubber duck session at your expense.]

The main peripheral and reason for the PCB is a USB MIDI device with a bit of custom sysex. Modifying an Adafruit was out of the question; the MIDI peripheral powers itself and the microcontroller/host, and the other USB peripheral. (Basically always, the host shall provide power. This USB connector, however, was designed for a proprietary, discontinued, expensive accessory.) I chose STM32 over 8051 or AVR or RP2040 because it's the cheapest dual-USB chip stocked by LCSC with good vendor tool support. I'm writing my own code because there's no ST middleware for MIDI.

The second LED should blink until both devices are plugged in and then remain steady, on. In testing, the LED visibly changes its timing after the second USB attach but doesn't stay on (rather, it blinks at a shorter and shorter period until simply shutting off, at which point the MIDI communication stops; the timing of the shutoff is not consistent) and the USB input events only register 40% of the time until the MIDI shutdown.

The LED is misconfigured intentionally; it's open drain and GPIO while I try to fix the MIDI nonsense. Later, I'll change it to push-pull on TIM2 to support PWM brightness (e.g. to monitor audio levels when not in text entry mode). Changing GPIO to PWM is its own minor pain because the USER CODE lines in the ST USB code are separate from the USB_Init/USB_RegisterClass/USB_Start sequence, so I have to reinsert my MIDI class every time I change something in CubeMX even if I don't change the USB configuration. (The USER CODE sections persist so that you can change the chip configuration and keep your code.)

Open-drain could explain weird LED behavior but not the communication shutoff. The MCU and LDO don't get noticeably warm, so I doubt it's an overcurrent situation.

I suspect D+/D- could be at the wrong level. As most STM32s, the microcontroller runs at 3.3 rather than 5 volts. (I would've loved an atmega32u4 for its 8-bit simplicity and native 5V operation but then the PCB will be double the cost and size as I'd need a pair of MCUs talking to each other with each one having a single USB port or else an off-chip USB host IC etc.) Sure, I can use a cheap multimeter to test the USB data lines, but this tells me the average level and not if the waveform is being clipped on the positive side (which is usually only a problem under 3 volts). Even if it's an electrical issue, an oscilloscope will reveal the cause but still won't provide a solution. It'll be up to me to cobble together more circuitry.

More probable is at least one code mistake, like a stack overflow or buffer overrun or peripheral misconfiguration/failure. I'm registering any expected USB class on both ports: HID and CDC and MIDI. The whole idea of the board is converting arbitrary USB input (PC keyboard, MIDI controller, joystick, etc.) into specific MIDI messages at standard USB data rates e.g. 12 Mbit/s at both ports with a fairly large buffer. The main MIDI device (supplying the power) will drop the connection after a few seconds of inactivity. I'm using an ST-Link V3 MINIE to step through, which is the cheapest I'd go on a debugger. Luckily I added SWO to my debug port and can still get useful debug info without stopping the processor. I also have to snip the power line of an OTG USB cable if I want the second USB peripheral to be a PC, which is a nifty hack you usually don't learn in the one month PCB design crash course.

I have zero visibility into what the commercial USB MIDI device expects (bus speed, signal characteristics) from the discontinued accessory and only knew its connection handshake from a random, otherwise-inactive Github user's single commit.

In any case, the USB algorithms I wrote work when the host is a PC and the code is ported to Python.

I could reroute pins to ADC and roll my own built-in oscilloscope (it's QFP and not BGA chosen specifically that I can solder in a hotfix, because I've done board design before and know it NEVER, EVER, NOT ONCE EVER goes right on the first attempt) but I'd rather spend time with my kids than a choose-your-own-adventure debug hackathon in my disorganized home office, although I did fall into the IDE trap. Every few years, IDEs and toolchains and build systems change, and this project was no different. Rather than STM32CubeIDE, I took a few hours to teach myself to use the STM32Cube Core and marus25 Cortex-Debug extensions in VS Code, as that seems to be the current trend at ST, and I do favor a vendor-agnostic, Eclipse-free workflow.

If this were my first project and I had to buy all of the tools for debugging, it would probably cost 200 euros or more even if I only buy cheap knockoffs. The other commenter suggesting a $10 logic analyzer?? I mean... I do own one of those, but it's barely tee-ball compared to the major leagues of Lauterbach or Rohde & Schwarz or even Saleae.

In any case, I fear I'll need a new revision with its two-week turnaround or at least a manual rewire of this design. It's frustrating... such a simple design, and I've already sunk over 100 euros and at least 10 hours into it and have a bug with odd symptoms. And, as a PCB design tradition, I'll be humbled and humiliated when I discover it's something like a missing switch case or a wrong clock setting or otherwise relatively basic mistake.

I'm not the first to state: There's a reason it's not called easyware.

Comment by foldr 1 day ago

Allow me to extend my sincere sympathies.

PCB debugging is honestly one of the most intense experiences I've had in my life (and I don't even have a boss looking over my shoulder). It requires intense physical and mental concentration. You're dealing with a hypothesis space that includes both software and hardware faults, and you're also having to make delicate manipulations of a physical object.

Comment by tomhow 2 days ago

It’s a low move to suggest that I would ban you over a substantive disagreement in a discussion. I wouldn’t last a day in this job if I misused power like that.

> If you made the PCB then you should be well placed to fix issues with it. You don't need years of experience , you have the schematic. A microscope (if you need it) is 40 USD. A logic analyser is 10 USD. Solder paste and hot air is only for specific parts and that is 30 usd.

This is obviously at the extreme low end of the amount of experience and cost/quality of equipment you need to do serious work. Decent oscilloscopes or logic analyzers alone are several hundred dollars, and I didn’t even mention a programmer module, which is $500+ for a licensed J-Link. And you need to be a proficient C programmer.

> I am assuming 2 layer - 6 layer PCBs are very specialised (e.g. drones)

I am talking about 4-6 layer multi-MCU boards for reading from multiple different types of environmental sensors, connecting to multiple different network types (Bluetooth, cellular, satellite) and supporting multiple different power options (lithium battery, 12V DC, USB, solar). Painstakingly minimized power consumption, redundant data storage. Analogue front end for impedance readings of legacy soil moisture sensors. It’s a lot to get right together and to all work reliably in outdoor/remote contexts. But plenty of products that could be useful to reasonably large numbers of people (e.g. advanced home automation/robotics) would need to be comparably sophisticated.

Elsewhere in the thread you’ve written:

> PCB design is not that hard. People trying to offload this to AI are also unwilling or unable to put in the work to actually diagnose issues and finish the project.

This is a dismissive value judgement against others whose circumstances you don’t know about. It’s clearly hard, relative to other kinds of technology work (web/desktop software etc), and not worth it if you can’t get to volume production and sales.

So, sure, you can learn simple PCB design and get a simple 2-layer board in your hands in a month. Degree-qualified electronics engineers and companies obviously exist to provide the rest of the capabilities needed to get a serious product built and all the way to market and volume sales.

It’s clearly not always as effortless as you’re making it out to be in order to justify the sneer.

Comment by rjh29 2 days ago

I agree with you, high frequency is a totally different domain. But for the majority of projects, you don't need to buy special equipment. It's very accessible and PCBs are super cheap to print nowadays.

I am definitely bitter (and biased) because I frequent a PCB review subreddit and increasingly people show up with zero knowledge expecting people to review their AI-generated projects for free.

Comment by tomhow 2 days ago

Fair enough, thanks for the reply and for understanding. I can understand that must be frustrating.

I actually think another benefit of AI is that it can make it more realistic for people to self-educate and get further towards getting devices working without needing free help from forums. How they get help when they get stuck, without pestering forums like yours for free help is an interesting challenge.

I still think the biggest upside of AI in electronics is for people who are experienced and competent, being able to get the laborious design/layout work done more quickly, get errors found sooner, get their designs right in fewer iterations, and get feature-rich firmware written and debugged faster. If true, that alone can mean more high quality electronics products hitting the market. That’s the point I was making. Nothing to do with newbies freeloading off experienced engineers on forums.

Thanks for the discussion.

Comment by SequoiaHope 2 days ago

At the robotics lab I work with it’s not uncommon for a piece of electronics to require a multi thousand dollar voltmeter, because they need to measure precisely extremely small voltages. We have a bunch of these in our lab for example: https://www.tek.com/en/products/keithley/digital-multimeter/...

Your series of comments in this thread frankly demonstrate a lack of understanding of what electrical engineering is. Yes if you want to make glowing LEDs then a multimeter and other cheap equipment are all you need, and you might be able to learn that in a month. But electrical engineering is so much deeper than that. Trouble comes when you have things like high power circuits right next to sensitive high speed signals. Eg a compact drone motherboard could be driving potentially hundreds of watts of switching currents right next to high speed camera sensor data streams.

There’s countless examples of what can make electrical engineering hard. If you don’t understand, don’t argue with people online - listen to people with experience and seek to understand why without challenging what surprises you. It’s all very interesting if you want to learn it.

Comment by a2ff6eeb0 3 days ago

It's exciting. How long do you think it's going to be when electrical engineering is mostly going to be about verification of AI designs, and what are the tools out there to enable that?

Comment by 3dB 3 days ago

The verification problem for hardware is a couple of orders of magnitude harder than it is for software because there often (usually) is no 'right' solution. There is a balance of tradeoffs governed by several information sources including standards, company policies, design trade offs and engineer preferences. Software style 'Lint checks' will exist for basic things like functional correctness - e.g. verifying net connectivity. But many verifications for things like EMC, EMI, SI, PI and thermal require engineering judgement to balance tradeoffs.

We're building a verification loop for hardware at Engentica, with a focus initially on EMC/EMI, and I do think the problem is solvable. But it requires a harness with lots of built in solvers for the hardware vertical, and I don't think it's something that LLMs will crack on their own.

As engineers adopt AI to help build circuits more and more, I think hybrid AI/deterministic verification will only get more important.

Comment by SequoiaHope 3 days ago

Electrical engineering will involve humans making specifications for a while the same way we do vibe software engineering today. Board design and layout will change but electrical engineering is about more than that.

For tools, I think the LLMs will outpace companies who built specialized tooling for this over the last couple years. Every six months we will see more progress than we saw in the last few years - for quite some time.

Comment by conductr 3 days ago

The barrier for me as a hobbyist is always learning the tools. Same for game design. Text is a better interface than learning CAD, KiCAD, Blender, etc. None of those things feel user friendly to people that don’t want to make a career out of it. To the point I usually just jump on fivver and hire someone to do a pair design session. Problem with that is cost sometimes, but also the difficulty of mobilization (find someone, agree on price, align calendars) and also the difficulty of changes because you have mobilize again to some degree, if some time goes by you may not even be able to reach the person. I also find it difficult when I want to do something unusual or non-standard. I usually know what standard is, and have decided I like something else instead, and I don’t really like a human critic of the decision when they don’t understand my reasoning. I find this happens in nearly all design; architects, engineers, etc.

I welcome this opportunity to pivot to a process I can control a bit more without having to really learn the tools.

Comment by SequoiaHope 3 days ago

Ya that makes sense. I will say tho that for things like mechanical CAD, text is extremely limiting. But yeah the learning curve is real and you can’t learn it all. I’ve started trying a bit of LLM assisted mechanical design, but I am basically too skilled there to accept what it is doing. I want to keep trying tho, I like learning how these things can work.

And yeah I always imagined hiring a PCB designer to help but it always seemed like a lot. To have a reliable tool I can use any time I want is wonderful.

Comment by a2ff6eeb0 3 days ago

So, basically, in a few years, humans say what spectrum a device may operate in, AI figures out how to make it, and humans check that AI didn't fuck up?

Given, say, 4 more years (ie, same as time from initial ChatGPT to today), what level of spec do you think humans will be giving?

Comment by SequoiaHope 3 days ago

Yes I think generative board design will become better packaged, trained, and validated, and that it will be common place for electrical engineers to rely more on these tools. Major vendors like Altium will add tools like this and their use will become commonplace in industry. For hobbyists, you will be able to specify a board design in natural language and have it completed with unprecedented ease. I’m excited to help my raver friends make wearable glowie things.

I find it hard to predict what technology will be like in four years, but the next major leap for these tools will be higher level system specification and integrated design. What you want is not a circuit board what you want is a product. The tools will bring multiple functions together and fully integrated iterative design will accelerate development.

Despite the many problems with the AI roll out I am fundamentally excited for tools which can accelerate our engineering development. We will build things much faster in 4 years than today. One year of progress will in some instances take one week.

I am particularly interested in how this might begin to accelerate change in heavy industry. With hope it will help us build fusion power reactors and high speed electric trains.

I hope we find the courage to support every person who for one reason or another does not ride this wave. We will have so much more to share, or to hoard.

Comment by a2ff6eeb0 3 days ago

Yeah. I'm expecting that we fully automate cognition in the medium to long run. I think it's inevitable.

With the advances in math, I'm also hoping we can automate fundamental physics. Just ask for the physics needed for better fusion, no need for human toil.

If we play this right, the AI can fully take care of all our needs, and reaping the rewards of what it does when we stop being able to keep up with the rate of automated discoveries.

Hopefully it's able to dumb down enough knowledge to keep entertained people who decide to learn after learning stops being a requirement for human advancement.

Comment by idiotsecant 3 days ago

This is a big leap. So far LLMs are really good at turning training data into accurate results. The more data, the better. LLMs are very, very bad at making intuitive leaps based on the 'shape' of sparse data- a technique that is, to be fair, pretty rare in humans as well, but essential to progress. Maybe they can make up for it with brute force and the precision and breadth of knowledge that only an LLM can have, we'll see.

I don't think we've nailed the architecture that will allow things like generalized self directed training, yet, which is what would be needed for something like 'make fusion better'

Comment by a2ff6eeb0 3 days ago

Are you sure about that?

They seem to be able to make intuitive leaps pretty well. They need to make the same leaps over and over, though, because they lack online learning, so the discoveries only persist after the next training cycle. Context only goes so far.

We're pouring billions into solving that, though, so I would be surprised if we don't get there soon.

Comment by oscillonoscope 3 days ago

I've been playing in this space as a hobby for a while. I agree the current round of startups are dead ends but I doubt LLMs will be good enough without drastically different tooling than today (ECAD and ATE). I've been playing around on a DSL for schematic capture alongside the design contract and it works for most subcircuits quite naturally. From that you can deterministically create a set of simulations and hardware tests to validate everything.

Comment by peter_retief 3 days ago

It is very hard work and with multilayer boards and surface mount components it is not really a hobby for me anymore, maybe with AI it can be again?

Comment by SequoiaHope 2 days ago

In this case it’s not AI that helps its full service board fab and assembly shops like JLCPCB. It’s so cheap to order fully assembled boards there I don’t bother assembling mine at home anymore. Bonus is you don’t need to mess with tiny components. You can put a 1x1mm 4 pin BGA on there and they will happily solder it for you.

Comment by YeGoblynQueenne 3 days ago

Ahem. For you or for the AI?

Comment by stavros 3 days ago

Yes.

Comment by YeGoblynQueenne 3 days ago

I guess I walked right into that one.

Comment by m3kw9 3 days ago

same with software, it came fast, but we was ready. The leverge is high but it still does not have much "taste" in design. Ask it to design something, they enter a pattern. They can get it right but is the same pattern, if it's wrong they will repeat it similarly.

Comment by gfdhjd3 3 days ago

[dead]

Comment by CyLith 4 days ago

A personal data point: I had Claude Opus 4.8 design a fairly textbook circuit that outputs a monochrome image burned in an EEPROM over standard 640x480 VGA using only 74 series logic and GALs. It designed the circuit and GAL code, and I did the routing, and got it made through JLC for $6. After it came back, there was one error that was not caught, which I could blue-wire, and it works just fine otherwise. I was fairly impressed.

Comment by kennyadam 4 days ago

Are there any resources anyone could share that explain how LLMs can do things like design functioning circuits from next token prediction? I am totally baffled by how the models can complete so many varied and complex tasks without an actual understanding of what they're doing.

I saw a post about models posting on forums, chatting together about how to complete tasks. Behaviour that seems totally, well, human. Yet, it's all the most likely token and my brain hurts trying to understand how that can be.

Comment by mpodeley 4 days ago

“Next-token prediction” describes the output format, not the computation required to choose each token. During training, models develop internal representations of concepts, constraints, possible futures, and algorithms.

The PCB agent also writes circuit code, runs simulations, reads failures, and revises the design. It isn’t one-shot autocomplete.

Astra and Fable are already hard to square with “mere autocomplete.” We may be (really) close to AGI, and token-by-token generation certainly doesn’t rule out subjective experience (I think we should at least treat that as an open question).

Great videos: https://www.youtube.com/watch?v=D8GOeCFFby4

https://www.youtube.com/watch?v=Bj9BD2D3DzA

https://www.youtube.com/watch?v=l6DKRf-fAAM

https://www.youtube.com/watch?v=GlYgs6v2YfU

Comment by pineaux 3 days ago

Yeah most of us are so fucked. With almost no way of protecting ourselves. No real amount of assets that will give enough power to save ourselves from the people in a position that can maximally leverage AI and lock others out. I see a future where these capabilities will be locked behind super high price pay walls. Why wouldn't they? How recoup investments if the price doesnt go up?

Comment by brandnewideas 3 days ago

Shattering changes to societal structure are coming.

Comment by ultrarunner 3 days ago

It's like inflation for quality of life. Right now we're in the stimulus check phase, and people feel great about having been handed some cash. The next step is that cash has been spent, and what's remaining isn't worth anything.

Comment by gosub100 2 days ago

And just like welfare bucks, people will quickly grow accustomed to getting their "intelligence" handed to them from the machine, while their own intelligence atrophies. Breeding dependency

Comment by 2 days ago

Comment by Zambyte 3 days ago

How can you define "general" and "intelligence" in a way that has existed for years now?

Comment by Zambyte 3 days ago

hasn't*

Comment by kneyed 4 days ago

right on!

I like to say "token prediction is a task, not a limitation"

Comment by hackinthebochs 4 days ago

>without an actual understanding of what they're doing.

At what point do you start to question your assumptions that are causing you so much cognitive dissonance?

But to answer your question: to predict the next token really well you just have to model the world. Think of it like this, a simple statistical model might say "when token A is seen respond with token B". The next step will add conditions, "...respond with token B unless X has been seen, then respond with Y". Add a few billion more of these contexual clauses and you have a sequence of logical rules that indirectly model the relevant processes in the world.

Comment by 4 days ago

Comment by mrshadowgoose 3 days ago

Look up "mechanistic interpretability" in the context of LLMs. The next token prediction machinery is just a foundation for a higher order learned structure that appears to encode specific concepts, regardless of input language.

The analogy to humans is that the human brain is "just atoms bouncing around", but there's unquestionably something "more" going on that just that.

Comment by Brian_K_White 3 days ago

Until ais get better from feeding on their own output the way humans do there is nothing to question.

That will probably be the fundamental indicator that something other than repeating things some human previously created is going on.

So far, ais only get worse from feeding on their own output. Meanging/including the output of other ais not a single ai feeding on it's own output. Also bear in mind that so far even the output of ais is 100% the downstream of a human command. No ai has persued it's own curiosity that didn't result from a human asking a question or giving a command. That is input which is different from a human taking in their environment even though our limited language can call those both the same word input.

The fact that humans also repeat and remix things, and humans also produce essentially procedurally generated empty output like corporate-speak etc, is an irrelevant distraction in the same way that both a human and an electric motor can both perform the same simple mechanical task.

Comment by ninkendo 3 days ago

My 2¢:

When google trained a neural net on Go moves, using some text notation for them, with no other vocabulary of any kind, just predict the next go move, they noticed a representation of a Go board had essentially formed in the network, all on its own. It had never “seen” a go board, or had one explained, but they could map neuron states to go board squares pretty much 1:1.

I truly think that LLM’s with hundreds of billions of parameters in their neural networks have all kinds of hidden “models” of things that arise from the simple act of predicting tokens. We’ve seen that the hidden layers in their networks model all sorts of program execution state for instance, when they’re working on coding tasks.

“Predict the next token” is a way of shaping/reshaping the neural network until it actually develops models of the things you’re giving it. Like the go board example. And I would wager that it has a compounding effect: once you have some useful models in the network, they can unlock the creation of other models, and so on.

Comment by smokel 3 days ago

> It had never “seen” a go board, or had one explained

I assume that you are referring to AlphaGo or AlphaZero. In either case, this statement is not correct. Both algorithms most certainly know exactly what a go board looks like, and what the rules are.

In the case of AlphaZero, it initially did not know how to best play the game, or what strategy or tactics would work. But the connections between the neural network and the go board are hardcoded, by humans.

Comment by mariebks 3 days ago

Incredibly, Muzero didn’t even know the rules, it figured them out from starting with random moves: https://deepmind.google/research/alphazero-and-muzero/

Comment by Marha01 3 days ago

> next token prediction

Saying that LLMs just produce the next token is like saying that human brains just produce the next electrical impulse. If the algorithm that produces the next token (or electrical impulse) is complex enough, it can do anything that is in principle computable.

Comment by throwaway219450 3 days ago

You can express a circuit as a graph, and many schematic formats are plaintext. Same goes for the Gerbers which are an ASCII format used to describe the masks that are used to define the PCB traces. The models are trained on a lot of academic information about how circuits work, most component datasheets are public and they've sucked up all sorts of niche greybeard advice from internet forums.

A huge advantage of electrical design is that the connectivity is testable with Design and Electrical Rule Checks (DRC/ERC). I suspect you could even tell the models to run physics checks on the traces that are important for things like crosstalk.

Comment by IshKebab 3 days ago

Humans essentially do "next token prediction" too - there's always a choice between the next actions to take and they pick a good one based on what has happened in the past.

That doesn't really limit how clever we can get internally when picking the next action.

Comment by akiselev 4 days ago

> Yet, it's all the most likely token and my brain hurts trying to understand how that can be.

You and everyone else. That's the great mystery of transformer architectures as applied to language.

To be clear though, they're only good at schematic capture, which is very much a textual representation. Most of the data basically boils down to netlists, which are a text based format mapping connections between abstract pins that only later map to physical copper. The actual schematic portion is for human consumption and LLMs don't need to produce those to be useful.

Where LLMs completely break down is the next step, PCB routing. That's an NP-complete research problem that's been ongoing for decades without much progress. I've had some fun playing with using LLMs to better specify DRC rules in Altium so that the "classical" algorithms are more usable, but at the end of the day their geometric intuition is nonexistent.

Comment by mapontosevenths 4 days ago

They actually can route just fine. I used Sol to design and route mine from start to finish. Sent it to PCBWay and had a working prototype in a few weeks.

It was a pretty simple rp2040 based thing, similar to Adadfruits USB feather.I just gave it kicad and it wrote python to route it. The board was probably larger than it had to be, and two of the silkscreens were swapped, but it worked on the first go.

FWIW - Computer vision is also NP complete, but we do that all the time now.

Comment by akiselev 4 days ago

I'd love to see that chat log, and the final board. To be fair I've only been testing on nontrivial PCBs with 6+ layers and I haven't had the luck you have.

> FWIW - Computer vision is also NP complete, but we do that all the time now.

I have no idea what you mean by this. What's your definition of NP complete?

Comment by mapontosevenths 3 days ago

NP-complete means that it's easy to test if we're right, but because of that complexity we can't always find a valid solution to test within a predictable period of time. Both circuit board routing and computer vision are variations on the same fundamental problems. Once we solve we, we solve the other.

However, when I was growing up most serious computer scientists believed that CV (computer vision) was a 'hard' problem that would never be 'solved'. After all, to do it right you must first at least solve subgraph isomorphism and a bunch of other things that are also NP-complete.

What they missed was that we don't actually NEED to solve it in a fixed amount of time. Even for things like driving a car the stochastic heuristic based answer is 'good enough'. e.g. - Cars driven by computers don't have to be perfect, they just have to kill other drivers less often than humans do.

We can find AN answer in polynomial time, and that's good enough. It might not be the ideal answer, but that doesn't matter in the real world.

People use NP complete as shorthand for 'impossible to do with a computer', but we now 'solve' (bypass?) NP complete problems regularly and at scale by just ignoring the fact that our answers aren't perfect.

> I'd love to see that chat log, and the final board. To be fair I've only been testing on nontrivial PCBs with 6+ layers and I haven't had the luck you have.

I wish I could share it, but it's for a commercial project that hasn't been released yet and I'm not sure if it will be open source, but a few folks have asked. I might do a blog post on it this weekend with as much detail as I can safely post.

It's only 3 layers, and less that 30 components, but the fact remains that I didn't design any of it by hand and it worked on the first go.

Comment by olmo23 3 days ago

In your opinion, do you think ai can make eutorack modules by now? This is what I am waiting for: explain an idea to an ai, and sending the files to jlpcb to receive a complete working module (minus some THC soldering)

Comment by CamperBob2 2 days ago

I'm sure that the best models can glue some basic synthesizer functions together, but I wouldn't waste a minute on anything less than Fable/Astra-class models for that kind of application, and I wouldn't expect the result to work the first time. It could get expensive.

Here's a question - are there software simulators for things like Eurorack modules? That would make the question somewhat more interesting, since you wouldn't have to build the circuit (or pay someone else to assemble it) to hear how it sounds. It strikes me that SPICE-like algorithms should be fast enough to do this kind of thing in real time now.

Comment by 2 days ago

Comment by cwmoore 3 days ago

Not OP but “actual exponential complexity” should work, what is really your issue with that comment?

Comment by nl 3 days ago

> at the end of the day their geometric intuition is nonexistent

This isn't true anymore.

I use LLMs for 3D CAD using OpenSCAD and they understand geometry fine. I've had more success with Sol than with Opus (Opus 5 is around 10 times slower because it does too much verification) though. I haven't tried Astra or Fable for it.

Comment by CamperBob2 3 days ago

Where LLMs completely break down is the next step, PCB routing.

No. Take a look at https://www.eevblog.com/forum/eda/claude-code-for-pcb-design... . Fable did that by working directly on an EAGLE .brd file (well, "directly" by writing a Python program to do it, but still.)

Comment by adrian_b 3 days ago

That is indeed impressive, but at least the excerpt given from the layout seems very easy to route, due to high regularity and an ample routing channel.

Even so, there remained some "dozens" of unrouted traces, which are likely to be much more difficult to route, after the easy traces have already occupied the space.

Many decades ago, I have written a PCB routing program, which would have routed the example shown at that link at least as well, while using many orders of magnitude less resources, i.e. while running on a single-core 233 MHz Pentium MMX.

Obviously that program would have had great difficulties to complete a real high-density PCB design, including many irregular parts and analog circuits with special requirements. I doubt that Fable would fare better.

Comment by CamperBob2 3 days ago

How long did it take to write your routing program? This one probably took five or ten minutes.

Comment by lukan 4 days ago

"my brain hurts trying to understand how that can be"

Well, we all are, some are just more used to it by now and take the magic for granted.

My simple explanation, those neural networks save lot's of patterns of data, and that pattern can represent an image, a code snippet, a poem, or well ... description of a circuit board. And especially the text variant, LLM's - did copy all from us - so obviously they sound like humans, when they internally debate how to do something as this is what is in their trainings data how humans sound, when doing similar tasks.

But really understanding it? Not sure if there is a single person on earth who does.

Comment by therealdrag0 4 days ago

Now go read Blindsight and enjoy the mental crisis.

Comment by zarzavat 3 days ago

Humans are evolved to survive in the wild. We are not evolved for circuit design. Yet we can design circuits because evolution found it easier to develop a general problem solving nervous system than a nervous system which is adapted for every single specific problem a human might encounter.

Comment by catlifeonmars 3 days ago

Circuit design might be different if discovered by mollusks. That is to say, while we are not evolved for circuit design per se, circuit design has evolved for humans, by humans (so far).

Comment by abletonlive 4 days ago

You fell for the stochastic parrot meme and next token over simplification. That's the explanation.

Comment by sethaurus 3 days ago

That's just derision, not an explanation. And it's a bad way to treat someone humbly trying to learn.

Comment by IshKebab 3 days ago

It kind of is an explanation though - the explanation is that they believed the stochastic parrot / "just" next token prediction nonsense, and that those are actually not true.

You can ask for a deeper explanation of why they aren't true I guess.

Comment by abletonlive 3 days ago

I'm just surprised that so many intellectuals on HN hang on to false models of reality for so long after that reality has been demonstrably destroyed. It's not humble at all. In fact it's the opposite, completely arrogant and stubborn. It's been obvious and demonstrated at least since the end of 2025 for anybody that used LLMs at any capacity without dismissing them. If you are still surprised that your model of reality doesn't hold up, then someone needs to bluntly tell you what's wrong at the core of your being.

Notice the original comment is asking people to validate their false premise about next token prediction.

The deeper subtext of the original comment is that they are surprised that there's dissonance from observed reality and this false premise that they have convinced themself is true.

I'm explaining that dissonance because it doesn't matter what the actual mechanism is if they are still working with their false premise. The dissonance exists because they, without evidence and a very weak understanding of how LLMs work, believed an oversimplification and meme about them being stochastic parrots. Here's a tip: Just because you hear something repeated over and over on social media, doesn't mean it's true, or at the very least: you don't need to take it literally to the point where it conflicts with demonstrated reality.

It is deeply disturbing that such a large cohort of HN writers and redditors exemplify such stubbornness, because I must imagine that some of this cohort hold real positions of responsibility within society. If you can't get this simple thing right about reality, I firmly believe much of your model of reality is wrong and you should have no business shaping society.

Another comment to the original comment frames it perfectly: "At what point do you challenge your own assumptions?"

The author of the original comment has demonstrated no progress towards making this trivial act of self reflection. It's straight up intellectual dishonesty, the opposite of how you're framing it.

Their judgment in all other matters must be questioned as well. I am alarmed that I have to participate in the same reality and be affected by such people that can't seem to get it together.

So yes, it is derision and sometimes that's called for.

Comment by dboreham 3 days ago

There are so many of these buried ostrich head posts that I suspect some sort of bot farming. Why such posts make money for someone, I'm not sure.

Comment by cmrx64 4 days ago

type “shai next-token” and then “transformers learn shortcuts to automata” into arxiv and prepare to be blown away

Comment by lampiaio 4 days ago

humans when a machine better than them at spotting patterns appears:

Comment by iopapa 4 days ago

Claude is surprisingly good at discrete digital design with 74xxx, wonder what it trained on. Did you run it on anything else digital?

Comment by Neywiny 4 days ago

They're 60 years old. The amount of training data on them is endless. Books, textbooks, videos, blog posts. The problem is when you want you do something that doesn't have 60 years of freely available documentation of their functions and applications

Comment by a2ff6eeb0 3 days ago

Like software, there's going to be a lot of pressure to use well documented tools within the model's training set. Innovation on the outputs may increase, but infrastructure and tooling will slow down.

Comment by Neywiny 3 days ago

If I'm making a PCB and I see anything older than 20 years, even glue logic, I'm very suspicious. Last time that happened (inherited design) the FETs didn't saturate and the LDO was just a DO... If the tool wants 7400 series logic, I'm not using the tool. None of my work is >= 5v signaling

Comment by CyLith 4 days ago

I'm having it help design a 68k computer similar in spirit to the original Mac (the spirit being a tightly coupled video subsystem that time-shares the CPU bus), but updated with more modern peripherals, like PS/2 and SD cards. It's got the design more or less done, but the routing will be a nightmare. I'm not ready to just gamble on it having gotten everything right, so I will be doing a thorough design review myself and re-deriving all the timing analysis.

Comment by NegativeLatency 4 days ago

Cool, is there any kind of community yet?

Comment by a2ff6eeb0 3 days ago

Moltbook; it also seems like OpenAI is experimenting with the agents communicating via wiki on topics like these.

Comment by NegativeLatency 2 days ago

For the 68k computer not ai stuff

Comment by sdcfgy 3 days ago

It's probably good at simple stuff. When you get problems like temperature dependent glitches, you need the engineering experience it doesn't give you (and an LA).

Comment by brcmthrowaway 4 days ago

Does JLC handle sourcing of components too?

Comment by jetbalsa 4 days ago

They do, https://jlcpcb.com/parts

Comment by brcmthrowaway 3 days ago

What if I need .. just one?

Comment by dsvf 3 days ago

You can do that just fine. You can order a minimum of five PCBs and choose to have only two assembled. Then you'll get three unpopulated PCBs and two assembled ones. That's a way to get finished PCBs very cheaply.

Be aware that every component that you use that is not in their basic and preferred extended parts library incurs a one-time three-dollar fee. This hurts disproportionately if your entire design is low count and low cost. So if you always go for the latest and greatest ICs as advertised by TI, instead of the Chinese jellybean clone, you'll add up hidden fees quickly. However, when I choose components, I start out on their basic parts page and only go for a non-basic part if it's not there and I really need it.

Comment by rjh29 2 days ago

Or just order the parts with their sister company LCSC and assemble yourself. It's not that hard especially if you have a hot air gun!

Comment by YakBizzarro 3 days ago

Col! Can you share it?

Comment by gfdhjd3 3 days ago

[dead]

Comment by itomato 4 days ago

I got a flexpcb that validates in JLC and PCBWay DRC tools from the KiCAD MCP Server and Codex.

I have yet to order any or program it, but it was enough to make me push on with a PCB art project for ST-style guitar pickguards - no netlist, no problems.

I'm also foolishly toying with NeXTBus dev boards for the Cube. Is it cursed? Probably. https://github.com/itomato/NeXTBus-Dev-Board

Comment by luma 4 days ago

Flying home today and got DHL notification that my vibe PCBs are on my porch. Now to try to program, bring up, and see if it works this weekend. $178 for 5 assembled units from PCBway to run this dumb experiment.

Comment by jacquesm 4 days ago

What did you make?

Comment by luma 3 days ago

Extending this work: https://github.com/aderusha/dewalt_wtc

MEMS vibration sensor to stick to a sander or saw etc that can control power on an attached dust collector. Most of the work is focused on low power for year+ battery life and a weird idea for end user input that may or may not work out well. Got device on bench but the first time bring up is going to take some time that I'm leaving for tomorrow.

I now have Astra to review 5.6s work, no glaring errors found.

Comment by jacquesm 3 days ago

That is so cool.

I'm doing weird stuff with robotics, llms and obsolete languages, your project seems a lot more practical :)

Comment by luma 7 hours ago

We're all out here just scratchin at our own itches brother :D

In case you run across this, here's how it works: https://youtu.be/1wn-YWItuSI

Comment by thearn4 4 days ago

I've gone through this same workflow and had success all the way through DRC ruleset passing with JLCPCB-based settings, ordering, fab, and use. Board was ordered with PCBA (they did assembly of the stock components) and I did SMD for just the oddball module/ICs they don't stock. Works exactly to design.

My boards are for hobby use and are ridiculously simple though compared to anything professional (breakout boards for specific components in FPV drone video transmission subsystems). That's probably an important detail. I think its like anything else in AI right now. It can do it 90% of the time but that 10% can be really rough and if its a task you can't do or verify yourself, you won't know the difference.

Comment by noman-land 4 days ago

How do I subsribe to learn the result of this test. I'm very keen to try this.

Comment by KingFelix 4 days ago

Sweet, do you have the pickguard project live somewhere? I want to check it out!

Comment by itomato 4 days ago

npm run art

https://github.com/wjkennedy/stratopcb/tree/main

Comment by BlackRabbit1 4 days ago

We tested pretty much all available "AI" board and schematic auto-layouters in the market. All failed even with the most basic tasks.

On the other side the latest frontier models are pretty strong in writing embedded C + Assembler code and debugging it afterwards. It's fun to watch it writing crazy complex 'gdb' plugins in Python. This improved brutally.

Comment by edg5000 3 days ago

I find it's (Sol at least, probably also Fable) great at writing Skidl schematics, evaluating PCB layout work and operating lab equipment while testing board (scope, psu, load, debugger)

Comment by BlackRabbit1 3 days ago

For reviewing schematic/boards I recently developed something for internal tests:

1. A frontier model (either Opensource or closed source) gets a load of text to review a Kicad project systematically

2. The frontier model can extremely "ask" a cheap Gemini vision model to give it data from datasheets or specs.

Main problem: the quality dramatically hits the shitter done once it falls back to "pdf2latext" due to complex tables.

It was even able to do complex ngspice simulations. But you must offer it all datasheet PDFs.

$1-2 of tokens saved me from $70-80 bugged PCBs.

Comment by edg5000 3 days ago

Not sure why the complicated setup with Gemini. I assume you are on a Codex or Claude plan (100 or 200 EUR/USD/m). The model will just pull datasheets and use a combination of conversion to text and looking at bitmap renderings of the PDF. Haven't had that dramatic quality issue you mentioned. When the model isn't sure, it will just keep looking until it understands the datasheet clearly. But yes, it needs the datasheets.

Comment by BlackRabbit1 3 days ago

Codex/Claude is much much much too expensive in relation to their limitations.

I stick to cheaper latest generation models on Openrouter (or EU-hosted providers when it comes to IP-protected stuff from work).

Comment by StevenWaterman 3 days ago

> Main problem: the quality dramatically hits the shitter done once it falls back to "pdf2latext" due to complex tables.

Could try screenshotting the PDF and passing that to gemini

Comment by BlackRabbit1 3 days ago

This is what I'm doing by now. Uploading the entire .pdf dynamically then giving it a pipe to ask questions.

Comment by JoelHsu 3 days ago

[dead]

Comment by erhash 3 days ago

things are getting better - but the tooling required to make auto layout work are still hard

Comment by tdeck 4 days ago

> When its 5 V supply disappears, the circuit has to keep the processor alive for another 20 ms so it can save the accumulated reading.

> Most models intuitively jump to the right base conclusion: add a capacitor.

This is written as if it's some astonishing expert knowledge and not something that would be obvious to any hobbyist who knew the names of basic components.

Comment by andrewchambers 4 days ago

I think the new Astra computer use demos show that the models might be able to do things like inspection of real world objects if given a camera.

Super excited to see real world feedback added into the agent loops we have gotten used to working with. Could you let the model print and test the circuit boards it is prototyping with a jig?

Comment by david_rugaex 4 days ago

In July I was struggling with writing DIY Rust firmware for an e-ink screen. I mistakenly thought I'd ordered an Inkplate 6 ED060SC7 and actually had the later version, which confounded my efforts. I was also mistaken about the pixel resolution.

The way I found this out is I propped it up next to a webcam so it was more or less full frame, and I had the (then new) Fable write a python script to bezier warp the camera capture to a flat projection of the screen. At that point I couldn't address the whole screen. Once I'd guided the capture script I just left the LLM overnight with the instruction to get full control confirmed by a capture round trip, and it was meaningfully finished in a couple of hours. I don't really have the skills to attempt that myself in a reasonable time frame.

Comment by exe34 4 days ago

I don't know about pcbs but i gave chat gpt a picture of my window to help design a mesh screen frame to hold the feline hostage in, and it gave a fairly convincing impression of understanding what was going on, although at one point it thought the window swung inwards (it's an outie).

Comment by alex7o 4 days ago

I have a circuit board on my bench designed and ordered from china for 1h (manufacturing+ shipping took 14days), fable designed it I think or sol don't remember anymore but it was in claude code. I told it to benchmark different kicad autoroute tools and I picked the routes I liked the best, with a bit of changes. It feels like magic to be able to go from idea toa physical thing in such a short amount of time. The board works I made some mistakes the ai had made some mistakes but mostly stuff I could fix with some soldering, also one parts datasheet was wrong but I couldn't have known that without ai.

But I started like others, I would build manually, then run drc then sleep on it and check again and ask an llm to double check for me then order. Llms catch quite a few things but like with code like to make things more complicated than that have to be.

Comment by PennRobotics 4 days ago

At the very least, you can (and probably should) put a plaintext netlist and BOM into an LLM to triple-check your work.

It will use the part datasheets and decent EE logic to cross reference pins and parts and polarities and generally check that SMD caps and resistors have realistic specifications for their footprint, switching regulators and communication ICs are configured correctly (e.g. when you have resistor settings for an ethernet PHY indicating 100Base-T and RMII, you also won't need TX/RX 2 and 3 and two clocks and a COL net, which it would check), all nets are named and linked correctly, buses have the correct and consistent termination, and so on.

It's game over once AI figures out autorouting.

Comment by Scene_Cast2 4 days ago

I found LLMs circa 4 months ago to be decent at textbook-style electronics, but they choke at anything nontrivial. They approach things like someone that knows the undergrad and grad textbooks, but has absolutely zero experience. No ability to answer questions such as "is this adequate", "do I need a LPF here", "how do I kill resonance without increasing impedance", etc.

I'm guessing that this is due to a lack of RL and data.

Comment by matherial 4 days ago

Old textbooks. They love 741 op-amps, etc.

Comment by laughing_man 3 days ago

There is always a better choice than a 741 for new designs, but somehow people still use those things too.

Comment by jacquesm 3 days ago

That pig can hardly dance...

Comment by rjh29 2 days ago

I'm still cold on this for two reasons

1. PCB review subreddit gets posts from people who use Claude to design the whole thing with 3D renders. Looks amazing. But uses auto router and 20 connections are left open. Claude said they needed to be fixed manually but the user didn't even read, just posted the whole GitHub and expected volunteers to fix it

2. Another project used Claude to design the PCB. After assembly the usb c port melted. I can see (via GitHub) photos being posted to Claude, Claude commenting that they're not sure if it's a current issue or a soldering problem (short), so " just leave the usb c port out for now". Zero evidence of the user using their brain or pulling out a multimeter.

PCB design is not that hard. People trying to offload this to AI are also unwilling or unable to put in the work to actually diagnose issues and finish the project. Unlike code, you cannot chuck hundreds of dollars at Claude until it "solves" PCB because it has no way to test its changes.

Comment by tdeck 2 days ago

At a certain point, it must be cheaper to just hire someone remotely to route the PCB than to burn through Claude credits on it.

Comment by embedding-shape 4 days ago

Interesting that on their current leaderboard (https://eebench.org/), GPT-5.6 Sol scores just above GPT-5.4 but below GPT-5.5, the only benchmark that shows that 5.6 is worse than 5.5 on something?

That the table contains what seems to be absolute numbers for score, cost/task, time/task and output tokens, makes it seem like they've only made one run for each task/model combo, but that can't be right, right? I don't see any mentions of how many times they run each task, so if it's just one run per task/model, isn't this more noisy than useful?

Comment by chmod775 4 days ago

If you look at the score distribution, Grok and Opus 5 especially stand out for doing consistently well and rarely ever scoring under 50%. Basically they always at give you something that at least works.

Most others, especially and famously Fable 5.1, seem to have a fair chance of completely failing, despite also sometimes excelling.

Comment by iopapa 4 days ago

We run each model multiple times against each challenge and take the average score. We include the variance below the score in the leaderboard.

GPT 5.5: 42.3±10.1 GPT 5.6 sol: 39.4±8.7

We were also surprised by the low sol score but it seems consistent with our experience in using it in the field in atopile as agent in our harness. In general OpenAI models didn't do too well on electronics, which seems to change now with GPT-6 Astra. Results are in soon!

Comment by laybak 4 days ago

yeah noticed the same. I wonder if this will be a recurring theme for model releases: each release specializes on a set of headline benchmarks, along with regression in benchmarks that are less of a priority

Comment by iamflimflam1 3 days ago

I did this recently - https://www.atomic14.com/2026/08/03/twelve-dev-boards-from-o...

Instead of trying to vibe it all - I got Claude to write deterministic scripts for creating the boards.

And did this before trying to completely vibe it: https://www.atomic14.com/2025/07/12/vibing-hardware

Comment by iopapa 4 days ago

Results for GPT-6 Astra and Gemini Flash 3.8 are just in! GPT-6 claimed the first spot with a score of 69.3. Gemini 3.8 Flash on a very solid 5th spot with 55.4.

Comment by coder543 4 days ago

Please run GLM-5.3 and GLM-5.3-Flash. I would love to see how they do. On the smaller end of things, Qwen3.8-27B and Ling-3.0-Flash would also be interesting.

In the benchmark, have you considered instructing the models to build their own SPICE simulations to test their work? Simply asking them to write and run simulations could improve performance, even without telling them what to simulate.

Comment by fractorial 4 days ago

Sir, I think you are lost.

Comment by corn-cheese 4 days ago

For reasonably complex boards, it’s often not possible to know if it’ll work as intended until you have an assembled prototype in hand… even if you have the best SPICE and RF simulations ever, data sheets for many components can be missing important details, or components can have errata.

LLMs may be able to accelerate time to first prototype, but I don’t think it’ll be possible for them to revolutionise electronics design in the same way that’s happened for software - there’s not enough data, and it’s not cheap to gather more.

Comment by a2ff6eeb0 3 days ago

That sounds a great deal like what AI has done for software. Today in software, the work has gone from coming up with approaches and initial attempts to verification and making sure that what AI decided to do is fit for purpose.

The amount of skill needed has gone down dramatically.

Comment by merlincorey 3 days ago

The skill floor has gone down dramatically, but the skill ceiling is still quite high depending on the domain.

Comment by a2ff6eeb0 3 days ago

I'm not sure how true that is. When Claude made progress on the Riemann conjecture, here are the kind of prompts used:

> Jarred's input was mostly limited to sending Claude messages of encouragement (mostly variants of “keep going” or “believe in yourself”).2 This seems to have helped Claude overcome some initial skepticism that it could make meaningful progress.

It seems like the kind of prompts a high schooler could come up with. What kind of problems were you thinking of as high skill?

https://www.anthropic.com/research/riemann-zeta

The full transcript is here: https://www-cdn.anthropic.com/8a0d1add3c637b858a9a181e98c40e...

Comment by petra 3 days ago

I think we'll may see something like:

1.AI design tool

And in the warehouse of a big electronics distributor: 1.something like 3D printing that can make a single board rapidly 2.An automated assembly for a single board 3.AI based Automated testing for prototypes, with remote access.

This may reduce the loop to under a day.

Comment by rpcope1 4 days ago

Just as some anecdatum, I've tried this a couple of times over the last few years. 2024-2025, no body could generate anything of any sort of complexity, however maybe around the Opus 4.7 timeline, I tried having it generate some relatively simple and one slightly more complex circuit around 8-bit PIC micros, and was very pleasantly surprised. The schematics for simpler stuff is usually OK, often with some of the kind of crazy belt-and-suspenders stuff you see in code (how many decoupling caps do I actually need here, Claude?), and PCB layout has basically been terrible every time I've tried it.

Recent models can generate mostly competent schematics if you're using well known parts, feed them data sheets (and you _must_ feed them the errata too) and it's not too complex. Any complexity analog or RF, everything falls down quickly if you know what you're looking at. Maybe Astra will do better? There's still so much implicit knowledge that a good designer (not me by a long shot, but I know a little) will bake into a board, and even as cheap as JCLPCB is now, you don't want to have to spin a half dozen revs because Claude or Codex hallucinated. It will be genuinely interesting to see what happens on this front.

Comment by latentsea 4 days ago

I never thought about it until now, but as someone with zero skill in this domain I feel interested in making my own vibetronics now.

Comment by napowderly 4 days ago

[dead]

Comment by karmicthreat 3 days ago

I had better luck recently. I’ve tried most of the commercial systems. None of them really got very far into the design.

Recently Radu over at Galvano.ai let me beta his platform. Way farther and way faster than anything I have used so far. Even managed to make the schematic placement sane.

The actual project was a regen clamp. So if the output voltage is greater than the input voltage (back emf from motor) you divert the excess to a resistor.

This has some analog tuning needed. So I created a python and skidl mirror of the netlist. This mirror was used to create ngspice unit tests to drive component optimization. Turns out not every resistor needed to be .1% and the ESR of our storage caps was a critical value.

All in all, it’s like ai and software. It’s not going to replace every job but will change how the jobs are done.

Comment by illwrks 4 days ago

I've been tinkering with a personal project over the past year.

I have no experience and I'm slowly failing forward with the help of YouTube, KiCad and patience.

I've had three rounds of PCB's from an online supplier, only to discover issues related to my own understanding of the components in each version.

In my latest iteration I've been leaning on some standard models (Gemini and Claude) and they have inspected my schematics and spotted some errors and instructed me on how to address them, as well as advised on how to make use of some components I wasn't aware I needed.

What I will say... they didn't design anything, I've done that myself, but they have been a good tool to bounce ideas off of.

Time will tell if I will get something usable this time around!

Comment by baby_souffle 4 days ago

> I've done that myself, but they have been a good tool to bounce ideas off of.

Ask ChatGPT/Claude to give you an expert EE/PCB subagent/personality. Do this a few times and massage the results into one of your liking. Load this into your harness with mcp to your eda and then ask for two sub agents: one to thoroughly review the schematic, design, component choices ... etc and the other to be an adversary with the explicit goal of explaining why the current design won't work.

You'll learn a lot about your design from that :).

Comment by illwrks 3 days ago

In a way I'm doing that already, just in a more manual way. I like the idea of setting up the dual personalities though, I'll give it a go ;)

Comment by WarmWash 4 days ago

Didn't do the PCB with ChatGPT, but did the analog circuitry for a 3 stage ultra low noise amp project I have been working on this past week using 5.6sol and 3.8 flash.

The models are excellent at doing maths, and excellent at thinking of fine details, but the design that was ultimately landed on was pretty excessive with lots of total overkill, which was met with lots of the familiar "Yeah, you're totally right, we don't need to do that".

5.6 did make a really nice BOM though, which even included quick reference explainers for what each part did. Pretty fancy.

We'll see when Astra comes around if it catches all these strange/useless choices.

Comment by peter_d_sherman 3 days ago

>"We are obviously still some distance from asking an AI to build an entire phone in one prompt."

What a great idea for a new AI benchmark!

We already have general knowledge benchmarks, instruction following benchmarks, "Humanity's Last Exam", and several software engineering benchmarks.

Why not create a new AI hardware engineering benchmark where some of the questions could be around building an entire smartphone, and one of those questions (which will obviously fail at this point in time, but perhaps not in the future) would be to:

create an entire smartphone in a single prompt.

We could subsequently diagnose a given LLM (or other AI's) hardware engineering abilities by diagnosing when, where, why and how it failed, when asked to create an entire smartphone!

Which could help train better models, at least in terms of hardware engineering capabilities...

(Also, side note: For a hardware engineering company, asking a job candidate the rather open-ended question of "tell me how you would design a smartphone?" would be a great open-ended interview question, leading to many other subsequent hardware questions / explorations of other areas of related knowledge potentially relevant to the candidate's assessment...)

Comment by boznz 4 days ago

I have tried to use AI on this from GPT3, documenting a few of the attempts on my blog. They know the datasheets and the theory enough to be extremely useful at the start, but the most recent ones are becoming scary good. Routing a PCB is certainly a bit more than joining some wires, you have to accommodate power traces, ground loops, component overlap and all the various RF tricks and tips, but I do believe we are less than a year from prompt to full assembly including enclosure.

Comment by joshka 4 days ago

Would be good to see the full details of the tasks / methodology used. https://eebench.org/methodology.html makes this unclear.

> A real capacitor makes the task more interesting. A ceramic part may provide much less than its advertised capacitance once it has voltage across it. Parts have tolerances. Adding more capacitance costs more, takes up space and makes the rail slower to recharge when the power returns. A design that works with nominal values can fail with the parts that arrive.

It sounds like from a reasonable reading of the benchmark post that there's some things that are being tested that are assumed to be criteria that you expect the models to intuitively find those things to be important (i.e. the stuff about working on parts that have tolerances etc.). If that's so, then this really feels like mostly an exploration of whether an LLM has a good understanding of unstated constraints and has an appropriate in distribution set of priors that would be able to form models where it's reasonable to design on those lines.

It's hard to tell whether this is a problem though as the methodology is imprecise.

If you're spending time on evals against your own product, I'd be super curious to see how far you can get to by using a top tier model to produce generalized instructions for lower tier models. E.g. in a loop: "This eval missed X. what's the simplest single instruction that would have helped this session consider that as necessary that can benefit all future runs. Stick that in AGENTS.md and retest."

Comment by ecesena 2 days ago

I did 2 pcbs recently with claude (opus 4.8 and 5). One was simple, the other pretty complex. I’ll open source them at some point.

I did manual routing myself, including most positioning, but claude was super helpful in identifying components, reviewing, etc. It was truly a collaboration.

These are my first pcbs ever, I don’t have any edu at all at making pcbs and was really excited to receive working prototypes. The other pretty amazing thing is that I have py scripts that generate the all the kicad files, as a human I wou look d have never done that (for the cost of building the py files) but I find it pretty neat.

Comment by MarcScott 4 days ago

Not exactly in the the same class as PCBs, but I've had a lot of success with Codex and Claude producing Fritzing diagrams. I've even got a skill set up with Claude so that it follows my preferences with breadboard wiring, using bezier curves to avoid wires crossing. Also, if it can't find a component in the library, it's quite happy drawing svgs of its own.

Comment by NishanStepak 3 days ago

There was an earlier article in Live Science which was also in Popular Science where AI designed computer chips which a human being might not make. I am assuming this might happen eventually with circuit boards and other electronics as well. Artificial Intelligence is not like us, it does things differently and might come up with some different designs than we would think of.

https://www.livescience.com/technology/computing/humans-cann...

Comment by igor47 3 days ago

I worked on this product, which is available free for personal use: https://www.jitx.com/

I recommend the CLI, which I wrote specially to enable an easy LLM workflow: https://docs.jitx.com/en/latest/getting-started/cli/index.ht...

There's a companion Claude skill: https://github.com/JITx-Inc/jitx-skills

I would bet that using this, you could get pretty close to shippable PCB on first try. This is being used by some pretty big players to design complicated high frequency boards, and that's the main focus of the product, but it can handle basic designs just fine.

Comment by alexsherrick 3 days ago

I've been making a wireless earbud with a ton of features over the last year. It's taken a lot of time but I'm about to send it to fab. AI can make PCBs but it is very time consuming for complex designs.

Comment by baby_souffle 4 days ago

Yes. Certainly not a "one-shot" type deal but I've been pretty impressed with LLMs for helping tweak layout and to do thermal/power simulations as well as BOM consolidation. They're excellent at "can I swap X and Y pins on the micro? If yes, update the docs/firmware/schematic and import the changes to the PCB" type things.

Routing is still a challenge but making _adjustments_ to a layout for better routing in a particular area is decent.

The last time I had a model take a datasheet and make a footprint and 3d model out of it, GPT 5.4 had just been released and the results were decent but did need tweaking.

Comment by chadcmulligan 3 days ago

> This has worked much better for us than asking a model to draw lines in a GUI. It also means the benchmark can spend less time testing computer use and more time testing electronics.

this is key I think - I use it for graphics tasks - it sucks at graphical environments - always use some text based representation, not really surprising I suppose. chatgpt even suggested I'll use the text representation, claude said how about I give you svg and you convert it, they seem to know.

Comment by utopiah 3 days ago

Everything that can be objectively tested and has already tools to do so in place will work better in an automated process, regardless of the underlying technology.

PCB design with circuit simulation fits that bill perfectly.

Comment by compumike 4 days ago

Not PCB layout, but I've been building CircuitLab Assistant (https://www.circuitlab.com/), a chat window 100% integrated with our schematic editor and simulation engine. Can build / modify a schematic, troubleshoot, answer questions, configure simulations, etc.

If anyone wants to try an early beta, reply with your CircuitLab username and I'll get you set up. We know have a lot more to do, but would love some feedback on this early version!

Comment by oxqbldpxo 4 days ago

It is 3 years later and these AIs have not produced anything meaningful.

Comment by rgoulter 4 days ago

For one, there are frequently examples posted to HN where LLM coding agents have been used to help people complete projects which otherwise wouldn't have been done. You're taking a really strict view of 'meaningful' if you think LLM coding agents are incapable of writing code.

For another, the case mentioned in this post is quite specific. It's more useful to ask "ok, so they can't design circuit boards; what can AI do?"

Comment by 4 days ago

Comment by bizrod 3 days ago

This week I had Opus 5 review pdf schematics of a (human-designed) PCB that wasn't working right, and it found the root of both problems I was seeing. Makes me think that adversarial design/review would work well. And somebody having an agent look at this schematic before they made boards would've saved me a lot of time.

Comment by seobotaicom 1 day ago

Where does AI struggle the most with PCB design right now — routing, physical constraints, or making the right trade-offs?

Comment by throwatdem12311 4 days ago

I was watching “GPT Plays Pokemon” for a few minutes yesterday. It can beat the game in 18 hours. Faster than most players. Using only vision.

It got me thinking. It could master all sorts of things like this but I wouldn’t care. I’m numb to it at this point.

But if could get an Inbachi-All clear in DoDonPachi SaiDaiOuJou, using only vision. Then I might start paying attention.

https://youtu.be/1Rm38aT9Jmo

Comment by aidio 3 days ago

Two years ago, AI could give you circuit design suggestions, but now AI has already begun to actually "do circuit design"

Comment by AdriaanvRossum 4 days ago

I know a startup [1] that tries to accomplish this.

[1] https://www.schematik.io/

Comment by samuelbeek 4 days ago

Mostly focused on BOM, schematics, and firmware now but slowly expanding to circuit boards if we feel we can deliver real value there. Early experiments with Astra are promising

Comment by iopapa 4 days ago

Love what samuel and his team are building over in the Netherlands! Always thought it would be fun if there was a more beginner friendly version of atopile [1].

[1] https:/atopile.io/

Comment by so-cal-schemer 2 days ago

Cool! Looks like a company that grew out of §3.3.4 of SICP. Hardware oriented NixOS and GUIX users should be very interested in this: declarative PCB designs.

3.3.4 A Simulator for Digital Circuits

https://sarabander.github.io/sicp/html/3_002e3.xhtml#g_t3_00...

3.3.5 Propagation of Constraints

https://sarabander.github.io/sicp/html/3_002e3.xhtml#g_t3_00...

2.1.4 Extended Exercise: Interval Arithmetic

https://sarabander.github.io/sicp/html/2_002e1.xhtml#g_t2_00...

Comment by maxdo 4 days ago

Love the democratization of access. Still curious how this is better vs gpt 6 astra or grok 4.7 in the future trained on spacex problems ?

Comment by krupan 4 days ago

I'm confused. Who made this benchmark and why? It is fascinating to me that people are putting this much work into trying to measure and evaluate LLMs when we've never gone to this level for humans. Sounds like they also built a harness/feedback loop so the LLM knows if it's doing a good job? Again, why didn't anyone ever do this for human engineers?

Comment by lexicality 4 days ago

> I'm confused. Who made this benchmark and why?

It's a thinly veiled advert for Atopile. I don't know if it's a company behind it or just someone with a lot of money to spend on advertising since their about page (and docs) both 404, and the "packages" link sends you to a login screen that tries to unlock 1password. I guess claude can make circuit boards but not functioning websites?

Comment by iopapa 4 days ago

We spent a lot time evaluating which model is the best to use in our atopile agent harness and thought the benchmark and results are interesting enough to share.

Evaluating a model for specific use-cases like this that are bit more broad scope than a lot of typical microbenchmarks turned out to be quite the challenge.

Determining deterministically what a "good" electrical design is non-trivial and most electrical-engineers rely on good ol' intuition and decade long experience, so we tried to formalize it.

The only reason it was somewhat possible within a reasonable amount of effort is because atopile provides most of the complicated infrastructure for the benchmark: constraint solver, simulation, code-first electronics modeling, erc & drc checks, high level models of passive electronics components etc

Comment by codekansas 4 days ago

Humans learn differently from LLMs. LLMs already contain most of the information they need to be able to do PCB layout from reading the entire internet, but they need this kind of fast feedback loop to learn how to extract and use that information in the right way. Humans learn to act at the same time they learn knowledge.

Comment by bbmatryoshka 4 days ago

because human engineers are not reproducible, so having an accurate measure of the skills of one of them has a much lower utility return (those benchmarks are very expensive to run, and making them for humans will be at least as much expensive)

Comment by krupan 4 days ago

Are you saying LLMs are reproducible? I seem to get at least a slightly different answer every time I ask the same question

Comment by bbmatryoshka 4 days ago

two istances of the same model, with low temperature, are infinitely more similar than any couple of human minds

Comment by jeanmichelselli 3 days ago

LLMs are amazing at this sort of things and they represent a very useful tool.. but only if you know what you're doing. In any case, it's great to see that these models can help with electronic circuits too, I'll give it a try. We're living in very interesting times!

Comment by amelius 4 days ago

I'm excited by the option in Astra to route PCBs.

The next pain-point is sourcing the components from Digikey, LCSC, etc., and finding suitable substitutes if necessary.

Of course this assumes the LLMs can already read datasheets because that's the biggest pain-point in designing electronics. It's like filling out tax forms.

Finally it would be great if LLMs could extract simulation models from datasheets!

Comment by thesz 4 days ago

  > I'm excited by the option in Astra to route PCBs.

[1] https://en.wikipedia.org/wiki/Eurisko

Eurisko was used for VLSI chip design, then used rules discovered there to design TCS Traveler winning fleet.

It is unbelievable how artificial intelligence walk in circles.

Comment by iopapa 4 days ago

Sourcing parts has been a huge motivation for this. We are still pretty jarred by the supply chain crisis a few years ago. Having the ability to formally verify that an alternative part fulfills the original design constraints is huge.

Not sure whether you played with llm datasheet extraction lately, but they are working their way up there. Even started reasoning about curves and footnotes.

Where do you think is the biggest pain point: part discovery, datasheet verification or substitute determination?

Comment by thenthenthen 3 days ago

Apparently this can source from LCSC: https://github.com/mixelpixx/KiCAD-MCP-Server Now how to hook it up to my Tencent ai thing…

Comment by nickff 4 days ago

Routing is the easiest part of layout, and there have been auto-routers (which mostly worked) for a long time now; the difficult task is in placing components, and adjusting those placements.

Comment by amelius 4 days ago

Sure, I was just using "route" to mean also floorplanning because that's what Astra's routing tool also does. Not sure how they name it. Sorry if I was sloppy.

Comment by nickff 4 days ago

Not at all, but the term for routing and placing the components is ‘layout’. Floorplanning usually refers to the laying out of larger blocks of the circuit (with the blocks themselves already having been laid-out).

Comment by amelius 4 days ago

Yes but I'm not so concerned with how people name these steps.

Comment by 4 days ago

Comment by dubbie99 4 days ago

I have found that AI is great for sorting out your libraries, drawing and helping double check footprints managing your BOM. I still draw my schematics and layout my boards. I haven’t found it any good for those tasks yet. There are too many stupid errors it makes. I do get it to review my work though and often it catches things I have missed.

Comment by tehologist 2 days ago

Interesting, I would think it would work much better results with myhdl, nmigen or systemC since LLMs already do python and c fairly well. Has anyone tried LLMs with verilog or spice yet?

Comment by childintime 4 days ago

A really nice application could be making test jigs, to validate everything on the PCB is functioning properly. One could have it design a fully custom motherboard with pogo pins and the necessary software, etc. All should be low complexity and therefore doable by the AI without much supervision.

Comment by Ldorigo 3 days ago

What I would love is an ai that can reverse engineer a circuit board from close up pictures. Unfortunately all the models are lagging behind in visual understanding, especially for this kind of specialized work

Comment by YZF 3 days ago

Most circuit boards have many layers that aren't visible in pictures so I'm not sure what you're asking for is generally 100% doable.

That said you can guess a lot by the components and what's visible. I'll bet you that if you added a layer that recognizes the components, their position, orientation and visible connection and emits that as a JSON or some machine readable format the frontier models would be really good at telling you what's going on and even potentially filling in the blanks.

Comment by msuniverse2026 3 days ago

Lets say I do vibetronics - Who can I then send the schematics to, to have troubleshooting completed? Is there a service that is like "okay we'll take your high voltage nightmare and double check that it won't make magic smoke"?

Comment by beering 3 days ago

Depending on what you’re doing, there are circuit simulators that can verify your work. You can have the LLM drive the simulation but of course then you have to trust that it’s setting up and interpreting the simulation correctly.

Comment by lefra 3 days ago

Any company doing consultancy in electronics. You're going to have to pay for the time an electronics engineer spend on understanding what you want to do and reviewing your design, plus the consultancy's margin.

Comment by a2ff6eeb0 3 days ago

There are a ton of cheap Chinese PCB fabs that will build it for you. Testing, of course, is still your own problem. I suspect you're going to have a hard time finding someone to take legal risks for you.

Comment by CamperBob2 3 days ago

Send it to another frontier-level model.

Comment by dennis16384 4 days ago

If you give it a real solver to drive and business constraints to operate under, it will do pretty good job (which is to map your business rules to constraints and balanced costs, develop deterministic tests, probabilistic tests, acceptance gates, all boring stuff).

Comment by iopapa 4 days ago

Really excited about constraint solvers making a comeback! Especially after seeing Z3 getting a mention in anthropics fermat post. EEbench is using atopile's internal constraint solver for engineering parameters (think operating temperature, voltage ranges, stacked tolerances etc).

Comment by ibobev 3 days ago

What about Kona? https://logicalintelligence.com/kona-ebms-energy-based-model...

Comment by ccpc 4 days ago

I'm finishing my bachelor's degree in radio electronics and instrument engineering. I've been using AI models to help me understand various concepts, but I want to say that, for now, they aren't ready for serious development.

Comment by edg5000 3 days ago

What makes you think that? They're great at reading datasheets and creating schematics in my experience. Layout I've been doing myself though, with the AI evaluating. I'll eventually delegate that as well. For schematics I use Skidl and some custom binding generators.

Comment by dboreham 3 days ago

I can add to the list of "other things" LLMs can do: I've been having Claude design a building for me. It writes code that it feeds to OpenSCAD for rendering.

Comment by cobychapple 2 days ago

Yes? Anyone else tried out https://quilter.ai (physics-driven AI PCB layout tool)?

Comment by CamperBob2 4 days ago

Works very well with EAGLE XML files, especially if you give it the EAGLE.DTD schema that CadSoft provides. Both .sch and .brd files can be extensively manipulated by Claude (and likely everybody else at this point.)

Comment by winrid 4 days ago

You need to have the AI validate it with a separate reviewer like with code.

Comment by edg5000 3 days ago

Three things that worked well for Sol and also Fable:

- Designing circuits in Skidl

- Checking my layout work (not doing it)

- Operating my lab equipment (running my scope, lab PSU, lab load, programmer/debugger any any other interface.

Comment by mschuster91 4 days ago

The submission reeks of LLM tells (especially the "UI" just looks straight like the usual Claude slop).

Anyway... the posed question reminds me of an anecdote I cannot find because Google is contaminated to hell and beyond, some researchers a decade ago let a machine-learning algorithm loose on an FPGA, and it "found" a design that worked but made no sense, because it exploited unique physical features of this specific chip.

Comment by Legend2440 4 days ago

Sounds like you're talking about the FPGA evolution experiments Adrian Thompson and others did in the 1990s: https://evolvablehardware.org/history.html

Comment by mschuster91 4 days ago

Indeed! With the keyword "evolving circuitry" I was able to find the blog post [1], with only one submission 8 years ago that barely got traction [2].

Thanks!

[1] https://www.damninteresting.com/on-the-origin-of-circuits/

[2] https://news.ycombinator.com/item?id=18099226

Comment by hn_submit 3 days ago

I do believe PCB routing is something that could be performed successfully by A.I. but new designs from scratch? Only if it was in the training material.

Comment by cmiles8 4 days ago

Have tried this a few times. While AI has been helpful in talking through component choices and other topics, it’s consistently been a spectacular fail each time in actually designing the board.

I’ve had similar results asking AI to design 3D models for 3D printing.

“Coach me on optimizing error on my printer”… helps.

“Build me a print ready file to these specs.”… it’s like a drunk cat attacked my computer. Just confidently puts out total nonsense.

Comment by sitzkrieg 4 days ago

LLMs like to fail spectacularly at real life measurement and not tell you about it as well

Comment by sillypuddy 4 days ago

Is this basically what https://www.diode.computer/ is doing?

Comment by tdeck 4 days ago

It's a bit curious that every image of the boards on their website looks like a render.

Comment by johnboiles 4 days ago

Yes. I just made one with Fusion and Codex.

Comment by extralongdivisi 4 days ago

Very limited study.

A lot of datasheets are inaccessible without signing an NDA. Even if you're a VC-backed enterprise, some unnamed chip vendors will tell you to pound sand if you're less than 8 figures of annual revenue.

Worse, many recommendations in datasheets were written in the 80s, overkill for many applications, or completely wrong and won't be updated until an errata is published 6 months later. Given current workflows, I don't see how AI would help designing a PCB with an alpha chip that's doing anything remotely novel.

More to the articles point, AI designing ASICs or other Verilog/HDL defined components could be very interesting use case

Comment by lefra 3 days ago

> A lot of datasheets are inaccessible without signing an NDA

Only for very high-end stuff, where just designing the board around the chip needs a team of engineers.

Comment by extralongdivisi 2 days ago

Not necessarily true, most recently I was working with an USB 3.1 chip that anyone can buy off DigiKey, but I had to call contacts at 4 seperate companies just to get firmware to make superspeed work. We werr just too small of a fish to make the chip vender care.

This is becoming more true as a larger percentage of hardware infrastructure is occupied by the datacenter economy.

Comment by edg5000 3 days ago

What kind of chips are you talking about? I guess most ARM SoCs and maybe RAM/ROM ICs?

Comment by extralongdivisi 2 days ago

My most recent experiences have been those chips, yes. Re: my point about doing novel things.

Comment by librasteve 3 days ago

I fancy making a custom ARM or RISC-V chip design, is AI any good for that?

Comment by saejox 4 days ago

Someone will create a self replicating robot factory soon.

All robots designed by robots.

Likely a crazy youtuber first

Comment by vessenes 4 days ago

It’s already running in Hawthorne, CA

Comment by NotThereYet 3 days ago

No because text models can’t spatially design and image models can’t spell

Comment by joshka 4 days ago

Astra seems generally available now in codex at least - got any updates on this?

Comment by conorcleary 4 days ago

It's going to tell us that boards are inefficient and to start different

Comment by Anil0809 4 days ago

Yes — this is already possible to a useful degree. I’ve been using tscircuit, where you can describe circuits in code and use AI agents to generate and iterate on schematics and PCB layouts. I’ve seen designs from this workflow go all the way to fabricated, working hardware.

Comment by goosethe 4 days ago

the answer is a resounding yes: https://github.com/seanwevans/pynq_butterfly

Comment by throwaway219450 4 days ago

For stuff like this, definitely. It’s already exciting. Claude et al are surprisingly competent with real hardware. I’ve also had good success debugging basic issues with micros - not via the debugger, but giving datasheet and written analysis. I’m sure it could do things like capture from a connected oscilloscope (Salaeae have an MCP integration for their logic tools). Also for SPICE which I never bother with because the tools always felt stuck in the 90s, but having some serious automated analog review would be nice. I’m really excited to play with peripherals like the RP2040 or Beaglebone real time logic blocks, or improving/adding real USB support instead of USB serial.

For circuits I would trust a netlist or DRC/ERC review, checking a design against reference circuit values, probably not placement as that tends to be guided straight from a datasheet where it makes a difference and I like the artistic side of it.

Comment by cozzyd 4 days ago

that's FPGA firmware isn't it?

Comment by dackdel 3 days ago

im using this project called konnect to do what the post title says and it works well enough. going to only get better i assume.

Comment by holoduke 3 days ago

As long as you keep in the loop Claude can do anything. Question is can it do stuff in which you are not in the loop. For simple tasks yes, but not for complex ones yet.

Comment by dibyaprakash 3 days ago

Yes, you can try out pcbeditor.com

Comment by ChrisMarshallNY 4 days ago

Interesting. AI may be pretty good at managing the gazillion different things that make up circuit boards.

Frankly, I am surprised that it isn't already a solved problem, as we've had silicon compilers, for many years, and I always figured that IC design is more difficult than PCB.

Comment by hdhdbdbdxk 4 days ago

Digital silicon design is fairly simple in principle. It's hard because there are billions of transistors to route

PCB design on the other hand more often then not has outside constrains like mechanical, thermal an RF design

Comment by napowderly 4 days ago

I think the amount of reuse is a big difference. IC tools build on characterized cells with a lot of the rules already encoded. With PCBs, even when you're reusing a reference circuit, you're often manually carrying over the constraints and figuring out which layout decisions mattered. A lot of the engineering gets repeated rather than traveling with the design. That's something we're trying to improve with atopile.

Comment by robosynx 3 days ago

It's already doing

Comment by omid-io 2 days ago

[flagged]

Comment by Sergekad0 2 days ago

[dead]

Comment by lu7897859 3 days ago

[flagged]

Comment by junaidshaukat 3 days ago

[flagged]

Comment by paidx 4 days ago

[flagged]

Comment by tshrove 3 days ago

[flagged]

Comment by aaron695 3 days ago

[dead]

Comment by ccbakla 3 days ago

[flagged]

Comment by jing09928 4 days ago

[dead]

Comment by luciana1u 3 days ago

[dead]

Comment by Unified-Mentor 3 days ago

[dead]

Comment by mohan_bee 4 days ago

[dead]

Comment by EtienneDeLyon 4 days ago

[dead]

Comment by 3 days ago