[HN Gopher] Show HN: Roons - Mechanical Computer Kit
___________________________________________________________________
Show HN: Roons - Mechanical Computer Kit
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
Author : uncial
Score : 121 points
Date : 2025-05-01 15:57 UTC (7 hours ago)
(HTM) web link (whomtech.com)
(TXT) w3m dump (whomtech.com)
| proaralyst wrote:
| 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.
| uncial wrote:
| 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.
| shafoshaf wrote:
| All with a reference to the Rockwell Turbo Encabulator! Awesome.
| uncial wrote:
| 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.
| sriram_malhar wrote:
| Good God! This is marvelous!
| Johnbot wrote:
| This is really neat.
|
| Is there any use for something like a hopper that dispenses new
| marbles continuously?
| uncial wrote:
| 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.
| convolvatron wrote:
| 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?
| uncial wrote:
| 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!
| anthk wrote:
| 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
| uncial wrote:
| 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.
| anthk wrote:
| 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.
| uticus wrote:
| > ...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)
| uncial wrote:
| 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!)
| uticus wrote:
| 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.
| BlimpSpike wrote:
| 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!
| uncial wrote:
| 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.
| thenthenthen wrote:
| The first idea that came to mind was, how to trigger sounds
| with these functions! Would be a mesmerising synth! Well done!
| cuken wrote:
| 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.
| uncial wrote:
| 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.
| cuken wrote:
| 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).
| sneha_tamal wrote:
| 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?
| uncial wrote:
| 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?
| duncancarroll wrote:
| 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.
___________________________________________________________________
(page generated 2025-05-01 23:01 UTC)