Show HN: Roons – Mechanical Computer Kit | Hacker News Reader

Show HN: Roons – Mechanical Computer Kit

whomtech.com

203 pointsuncial30 comments

I built a mechanical computer kit: https://whomtech.com/show-hn

tl;dr: it's a cellular automaton on a "loom" of alternating bars, using contoured tiles to guide marbles through logic gates.

It's not just "Turing complete, job done"; I've tried to make it actually practical. Devices are compact, e.g. you can fit a binary adder into a 3cm square. It took me nearly two years and dozens of different approaches.

There's a sequence of interactive tutorials to try out, demo videos, and a janky simulator. I've also sent out a few prototype kits and have some more ready to go.

Please ask me anything, I will talk about this for hours.

-- Jesse

Open articleView on HN

This is really really cool! The physicality of it is special and can be a huge help with some people to gain an understanding of whats actually happening on the micro scale. Reminds me very much of "Spintronics", a game that holds a special place in my heart as I could teach a traditionally conceptual topic to my kids.

Are the designs you've come up with 3D printed? I feel like there's a huge possibility of community advancement into this ecosystem (fully appreciating you should make a return on all of your time and creativity).

Thanks again for sharing something so cool.

Thank you! Yeah, I was trying to lean into the idea of "shrink yourself down into a computer and physically manipulate the bits", think it's the best way to understand what computers are actually doing.

Spintronics is really wonderful, not just for its cleverness (which is extreme) but also the total concept and aesthetics -- absolutely something I'm aspiring to.

And yeah these are all 3D printed. Agree with your sentiments around community stuff, I don't have any fixed ideas there but I would be absolutely delighted to see how people can build on this. There are so many possible physical cellular automata to explore; this is just one.

If you ever open up the designs or want another set of eyes to print on some other devices / take a crack at generating new components, please let me know (looking forward to the kickstarter).

> ...the gears have a layer of phase baffles (I don’t know the technical term). These physically block the gears from connecting until they’re perfectly synced up...

perhaps the correct term is "key" [0]? only thing i could find to contribute to this masterful project, by pointing out unimportant details like this.

[0] https://en.wikipedia.org/wiki/Key_(engineering)

That sounds like the right general category, but maybe not specific enough? We could call it an Uticus Key and see if we can get the name to stick, that might be funny. (Thanks for the kind words!)

btw love the about page

> Do you want to work with a company servicing 6,000 Clients across 8 Different Countries, with turnover of more than 125 Million USD?...Are you looking for a Proven Track Record delivered by an Award-Winning Multinational Conglomerate with over Two Hundred Years of Business Experience?...If not, whomtech has you covered.

This is incredible, love how much further you've taken it than just the proof of concept as well.

This is probably just my opinion, but I kept thinking that a better word for "drive" given the domain would be "clock". It's basically a binary clock signal driving these, right? Maybe "clock drive"?

Yeah, I'm not 100% sold on "drive". There's a few competing ideas here -- initially I called it a "loom", but then I wanted to lean more into the analogy of loading a disk, so "drive" kinda won out.

I get the feeling there's some word in the vicinity of "clock" that I'm missing. Something like "metronome", "pulse", "synchronizer" etc -- because you're right, the (primary) purpose of the drive is to deliver phase information to the disks. Drawing a bit of a blank tho.

(Thank you for the feedback and encouragement!)

As a CS prof, I'd love to have this in my office for students to play with. Looks awesome!

If the "Turing Tumble" was Duplos, this is moving up to Legos!

https://store.upperstory.com/products/turing-tumble

That's exactly what I was aiming for! I always thought Turing Tumble was such a cool concept, by why not miniaturise it?

Funnily enough I'm currently building out a "Duplo" concept to make roons easier for kids. It's a disk with double-thickness bars (so a 4x4 grid instead of 8x8) but that loads directly onto the standard disk drive, so everything stays compatible. Still a WIP tho!

that is awesome, how do you see this evolving for practical use cases? Is it just for education and experimentation, or could something like this scale for more complex tasks?

So I don't think this specific implementation has practical use cases (someone prove me wrong!), but there's a really really cool general point here--

We've had this technology for centuries.

Seriously. This doesn't need transistors or clever materials. Mechanically, it's much less complicated than what (say) 18th century clockmakers were doing -- it's just bars going up and down!

So if you'd asked me 200+ years ago, I'd say: this device can compute nautical charts, calculate differential equations, and some third incredibly useful thing. Nowadays we can do all that much better with silicon etc, and I don't see this competing practically on that playing field...

... but I think it's useful mental technology to notice that there were simple ways to perform arbitrary computations, accessible much earlier on in the tech tree, that sort of got skipped for some reason. So while roons is probably siloed to education/experimentation/fun, I really hope it inspires someone to go -- what else are we missing?

This is cool! The simulator was useful for understanding what was going on, I hadn't realised until I watched a few that the roons can push marbles out in between squares.

Thanks! Glad to hear the simulator was worthwhile, took me ages to hack it into shape. I did actually experiment with adding markings on the edges of the path roons, so that when you put them together, the "phantom channel" becomes visible. Ended up looking pretty cluttered though so I scrapped it.

This is the coolest thing I've seen in a long time. I've always wanted to see this visualized and the marbles make it so tactile and real.

This looks like you make a shapez-like game and then make it runs in real world, too impressive for my brain to understand how it works!

How loud is it? Would it disrupt the whole office if I had it on my desk and occasionally played with it?

Very cool, I'll check back for the Kickstarter!

With the FDM-printed prototype version there's a noticeable "a mechanism is occuring" sound, largely due to layer lines. Tbh it's not too loud rn but I'm confident we can get it smooth and whisper-quiet using any vaguely sane manufacturing technique (defaulting to injection moulded ABS but we have some options). The marbles/roons themselves are silent, it's just a question of getting the barrel cams to play nice.

The first idea that came to mind was, how to trigger sounds with these functions! Would be a mesmerising synth! Well done!

you implied that there were scaling problems. it would be really fun and indstructive if one could use enough nodes to bulid a simple store and von neuman machine with instruction decode. but this seems like it would take a 10x10 array of your panels. do you think you can push this architecture that far?

Yes, fantastic question! 10x10 is actually way more than we need -- don't know the minimum but I suspect 2x2 is possible. The interactive tutorials sketch out an approach to this. My instinct is to use SAM with timers for memory/program, then there's a couple of approaches in the tutorial for the instruction processor. Toying with the idea of an OISC to minimise space requirements. Tho we can always add a motor to the encabulator if we need to drive a larger grid!

Would subleq work on this? Also, the 'ascii' output array/grid would be like a whiteboard, and a Forth on this would look like the computers from "Slouching towarsf Bedlam" text adventure.

Maybe it would need 5 minutes to show a prompt, but, if it works, it would be the greatest homage to computing (and Chuck Moore) ever.

There's Eforth for subleq, and the muxleq version feels almost like a native Forth on speed.

https://github.com/howerj/muxleq

The muxleq repo has the original subleq too. You can set the options for floating point and the do...loop (among others) in the Forth file and then you could just recompile the DEC file.

With the do...loop it's closer to the standards. And, yes, I can mimic for...next with ease with most ans forths such as pforth.

If any, I can link an already recompiled and more featured DEC file here.

BTW: You would like the book from the author too.

https://howerj.github.io/subleq.htm

I'm absolutely going to need a few hours to properly digest this, but here are my preliminary thoughts:

You could almost certainly do a subleq in roons! Tbh it may end up being a practical approach to making a compact full computer -- though seems like it might compromise legibility. This is way outside my wheelhouse, thank you for the links, I have a lot of reading to do.

Re: an ASCII grid -- I explore this a little bit in the tutorials, but because of the up/down nature of the loom, each marble only occupies 25% of the surrounding whitespace, so this kind of display doesn't end up very visible. This is why I'm leaning more towards the peripherals approach for rendering stuff.

Or just send the keys to a typewriter, as some guy did with a ZMachine, Asterisk and the mentioned device. But it would need a microcontroller, or a way to roll a ball in a precise way to type down a letter.

This is really neat.

Is there any use for something like a hopper that dispenses new marbles continuously?

Absolutely yes! I prototyped a hopper peripheral that doubled as marble storage. It hooked into the gear grid and used the rotation to drive a helical lift to carry the marbles up to the surface, where they'd enter the edge of an adjacent disk. I ended up setting it aside for 2 reasons: 1) Lack of time -- sad to say but I had to focus on getting the roons core ready. Too much to do! 2) The HDD peripheral -- this is a WIP that compactly stores and emits bitstreams. The way it's designed, it can be set to "reservoir" mode, which emits the continuous marble bitstream like you suggest. I figured it'd be redundant to have both a hopper and HDD, so I scrapped the hopper for now.

All with a reference to the Rockwell Turbo Encabulator! Awesome.

I'm grateful to Rockwell for their excellent research, but I think they really dropped the ball on side fumbling -- any competitive encabulator in 2025 really needs to be using bifurcated resonance.

Good God! This is marvelous!

Reply on news.ycombinator.com