Absolem Splash


I’ve designed and built my own mechanical keyboard. It’s fucking awesome! I’m going to ramble about it now, in detail. Read on if you’re interested, or jump to either the in-progress or the finished pictures, or post a comment to the reddit thread, or… you know… do whatever you fancy. I’m not your mom. :)



Until about one and a half years ago, I’d been happily typing on a basic Genius keyboard with a QWERTZ (Hungarian QWERTY) layout. Ah… simpler times!

I was hovering at about 50-60 wpm, which – while decidedly not blazing fast – didn’t bother me much. I also didn’t really care that my typing “technique” involved around 4-6 fingers and a lot of looking at the keyboard. What did bother me was nights when I couldn’t actually see the keys and it slowed me down quite a bit.

If you’re thinking that my solution was learning to touch type, you’re wrong! (for now…) I, of course, decided that I needed a backlit keyboard. Around this time I was vaguely aware of mechanical keyboards and the “supposed” superior typing experience they provide. So to celebrate my dissertation defense, – and after a cursory glance at full size vs. TKL arguments – I treated myself to a MasterKeys Pro S. And that’s where the problems started…

To be fair, the Pro S is a fine keyboard. But it’s not even the thing that convinced me about mechs. By the time it arrived, the geekhack-deskthority-r/mk Bermuda triangle sucked me in, and I was already too deep. Looking back now, I think the main cause was that I started looking at the topic as genuine “research”, and in my mildly fanatic1 worldview that could only end with another “dissertation”, which is what this post is, I guess.

So, strap in as I rant about the whole journey that led me here. Also, fair warning that I’m writing this on the already finished Absolem, which is just a pleasure to type on, so I’m going to be verbose! :P


Being a researcher by trade, I can very much appreciate the need for seeing what someone has already done in order to not reinvent the wheel. Also, following the old saying “stealing from one source is plagiarism; stealing from many is research”, I have basically patched together the (imho) best parts of what the current state of the art has to offer. It was an interesting observation to make that almost all the best2 ergo aspects came from different places, while their combination didn’t exist yet. That’s probably what lead to me deciding to design my own; had I found a board that checks all the boxes, I’d have just ordered that. (I’m also secretly happy that it didn’t turn out like that, because this way I got to make my own, and it was a good chance to grow… but psst, don’t tell that to anyone!)

So with that in mind, let me just quickly walk you through the steps that lead me to “keeb enlightenment”. Disclaimer, though: I’m only going to mention most concepts briefly to keep the post’s length manageable, but it can hopefully serve as a good starting point to begin your own, deeper research if you want.

General stuff

Okay, basics first, if you’re interested in the topic, you should browse GeekHack, Deskthority, and (of course) r/mk. (As an example, here are my tentative first steps in a brave new world over on Geekhack).

These places are not only chock full of information, ideas, and inspiration; they also house a very helpful and supportive community. This is also where I’d like to thank a few people for their general help in this project, namely:

  • algernon, fellow Hungarian keeb expert, for all the early advice,
  • Azel4231, for his help in switch layout related measurements (I’m the “redditor” from the addendum…),
  • and, naturally, DotDash32, for the metric shitton of discussion we’ve done in both posts and reddit messages that really helped me shape what I should aim for.


What became really clear early on is that row-staggers are evil. The reason they exist is pure path dependence (we’ve always done it like this, let’s keep doing it like this), and they should be eradicated. For me, this is an issue that’s been a non-issue for a long long time, but now (that I’m “enlightened”) it’s impossible to unsee… I mean, I sympathize with all the muscle memory that will be lost in a transition (I’m in the middle of one right now, after all), but that can’t be a good enough reason not to switch! I’m not really a comment-y kind of guy, but you can find even me sometimes in the middle of an argument under reddit posts that claim some connection to ergonomics, yet still retain the row-stagger.

I’m much more “lenient” towards ortho (a.k.a. grid, or matrix) layouts, but the clear winner of this aspect (for me) is column-staggered boards. Let’s give future aliens a chance to figure out how we looked like!

What’s more, I’m very much in favour of an “aggressive” stagger. Many boards started in the right direction, but few went far enough, so I’ve been planning to be a little heavier-handed in the stagger department from the start.

As an example, compare a “traditional” TKL (or 60%) vs. a Planck vs. an Atreus.

The number of keys

Today’s full size (and beyond) keyboards come from the assumption that there should be the same number of keys as there are desired functionalities and we should make our hands conform to the resulting layout. I, on the other hand, think that the inverse of this is true, namely that we should make the number of the keys match what’s comfortably reachable from the home position and make the desired functionalities conform to that.

This leads to a) touch typing – or at least a strictly enforced finger-key relationship (which has many more benefits I’m not going to discuss here) and b) to the need to significantly decrease the number of keys. On the other extreme of the spectrum is stenography, but even if we remain firmly within the realm of letter-based typing, we can (and should) make do following the “at most 1 key distance from home” principle. That leaves at most 6 × 3 keywells + 3 thumb keys per hand. I’d argue that anything more than that is bad. (Not only “unnecessary” or “wasteful”, mind you, but actually bad. As in, it could be better with less…)

The natural result of a small number of keys while still wanting a large number of functions is the use of layers. And layers – especially if combined with custom programmability, more on that later – are the “bees knees”! Nevertheless, I encounter many posts that criticize the overuse of layers, or posts that express confusion about how a 40% keyboard can still be practical. I’d refer the former group to their shift keys and the notable lack of dedicated capital letters on their boards, while the latter group should take a look at their phone while writing a text and tell me again how a really small keyboard is unusable.

Anyways, I digress… The point is that I came to the conclusion that there should be very few keys with heavy layering support. As an example, consider traditional keyboards vs. a corne.

The pinky column

Like we saw above, the pinky often gets 2 columns (similarly to the index finger) even when conforming to the “1 distance from home” (1DFH) rule; and much more when not. What I’ve found is that a) it’s unnecessary with a sufficiently clever keymap and a lo(oooo)t of practice, and b) it should be avoided to spare your weakest fingers however you can. So I’ve adopted a further restriction over the 1DFH to limit myself to 5 × 3 keys per hand (plus the thumbs, of course).

Regarding the physical layout of the pinky keys, my experiments (and my eyes, when looking at my hand) showed that the pinky can use a little bit of separation from the others. This could theoretically apply to the ring and middle fingers, too, but I didn’t feel the need in those cases. However, it really shouldn’t apply for the index finger, which already has an extra column to take care of, like in the case of the Sector.

As an example of pinky overworking, consider any regular layout (or even the Ergodox) vs. the Minidox (of which my design is basically a slightly refined, glued together, and wireless-ized version). As for an illustration of the pinky angle, take a look at the ErgoWarp.

The thumb region

Generally, there are three approaches for the thumbs:

  1. SPACEBAR!!! – one of the thumbs can keep hacking away on a button that takes up 6-7 spaces, while the other just exists. Very efficient… </sarcasm> There are more sophisticated layouts, with split spacebars and the like, but from an ergonomic standpoint, these are all subpar compared to the next two.
  2. Clusters – consider the Ergodox again. While this way is definitely better than a single spacebar, in my opinion it overcompensates with the amount of work it tries to give to the thumbs. The side effect of this is that very few of those thumb keys are actually convenient (or usable, according to some). This leads us nicely to:
  3. Fans – consider the KeyboardIO Model 01. This approach appreciates that the thumb actually moves in an arc, and doesn’t try to add extra functionality either above or below it.

This is probably the only “really” original part of the Absolem design, as all the other stuff I’ve mentioned so far could be seen somewhere before. And, depending on how we interpret “original”, maybe not even this… But: I actually placed the thumb keys on an arc, with a measured thumb radius. Yes, I actually had to refresh my trigonometry for this! And I, of course, followed the 1DFH principle, too, so there can only be 3 (unlike the KeyboardIO’s 4).

I’d also like to mention, as someone with quite wide (read: fat) thumbs, I’ve aimed to have 1.25u thumb keys from the start, at least for the home position. The sides can more easily be 1u because they don’t have a neighbor on one side, so there’s less chance for misclicking (mispressing?). But the thumb home position (which is flanked by other thumb keys on both sides) definitely deserves to be bigger imho.

The angle between the halves

Like we already established, our hands don’t just sprout out from our chests. The forearms (like the thumbs) move in an arc, if we consider the elbows fixed. That means that the keywells for the hands should be in line with each other (parallel? co-linear?) only if they are shoulder width apart, which they rarely are.

So for a “normal” compact board, no angle is definitely bad for the wrists. Splits solve this quite elegantly by completely relegating the issue to the user, but for non-splits there should be an angle. Additionally, my (very un-scientific and subjective) measurements indicate that that angle should be at least 20° (so a little steeper than anything I’ve seen, apart from crazy adjustable ones).

For illustration, consider traditional one-piece boards vs. splits like a Let’s Split vs. angled one-piece ones like – again – an Atreus.

3D aspects

What we haven’t touched on yet, are 3D aspects like forward/backward tilting, inward/outward tenting, and concave curvature within the keywell. The curvature was probably the first thing I abandoned because a) it prohibits a lot of manufacturing options that lead to an “heirloom-grade”3 keyboard, and b) it is really only useful for larger keywells anyway, which we’ve already ruled against. As for tilting/tenting and separate thumb planes, I gave them a chance and then decided not to utilize them, as I’ll talk about in the design section.

For examples of the concepts, see the Dactyl-Manuform which conveniently has them all.


To keep it simple (and to avoid straying into religious wars territory), all I’ll say is that generally there are linear, tactile, and tactile+clicky switches, where the former two can be silent or regular. Plus I wanted to stay within the bounds of the MX standard (mainly for keycap reasons), so if you’re interested in more extreme switch choices, I’ll have to disappoint… A round of research on the interwebs suggested that linears are mostly for gaming, plus everyone seemed to agree that a tactile switch is better for typing purposes. There’s also this speed typing guide that states that feedback is good! So linears were out.

That leaves silent tactile, regular tactile, and clicky. Now, if sound is important (which it is for me), then there is a really weak claim to endgame status for any regular “in-between” switches. You either want it to make sound (in which case you’re probably one of those clicky fanatics with MX blues in a workplace board :P) or you don’t (in which case you go for silent options).

You might have deduced from the wording of that last sentence that I went for the latter, and if you constrain your search to silent tactile switches, you’ll inevitably run into Zilents. There are cheaper alternatives, but since this was an endgame build for me, I went with the Zilents, and they are absolutely worth it. People might chalk all the rave Zealio/Zilent reviews up to choice-supportive bias, but I know I love them, so there’s that.

Another thing to note here is that I’ve seen very few boards out there that utilize variable spring weights for their switches. However, if we hold ourselves to the notion that each key has its dedicated finger, we could account for the different finger strengths by the different switch weights. So that’s what I did. Luckily I didn’t even have to take apart and modify any switches manually, as Zilents come in four flavours. See my ergo finger-aware weight layout in the Assembly section.


The aesthetics are discussed below; the focus here is the profile of the caps, which greatly add to the overall feel and ergonomy.

Side story: during my “awakening”, I went from staring at high resolution photos and not seeing any difference other than color and symbols, to being able to identify SA, DSA, XDA, OEM and Cherry by a cursory glance. This is probably not a very notable feat amongst the hardcore keeb folk, I just put it here to remind myself (and the esteemed reader) that it is easy to forget what is “obvious” and what isn’t to an outsider…

Anyways, I knew early on that I value sculpted sets, sort of as a compensation for not having curvature within the keywell. Plus I’d like to mention the “feedback is good” principle here again, which also points to sculpted. And from among the sculpted sets, I felt that lower is better than higher – so SA was out pretty fast.

Then came a research session from all over the net to see what people like/dislike, and once I factored in my priorities (mostly typing comfort, thereby largely eliminating gamer concerns) I found a pretty unanimous preference for the OG Cherry profile. Now I have to note here that I therefore don’t have personal experience with other profiles, just the stock OEMs of the Pro S and then the Cherries. But having tried them, I don’t really see myself going for anything else. Cherry for the win, case closed.

Actually doing something

The research section cannot be complete without mentioning how much reading and browsing I had to do in order to at least have a superficial grasp on how anything works within the field of electronics in general. I mean I haven’t held a soldering iron in my hand my whole life. And that doesn’t even take it into account that I’ve always kinda sucked at anything DIY4.

But that’s in the past now, and a PhD (plus a short but very intense job at a cutting edge tech firm) quickly teaches you that you’re capable of much more than you initially think. So even if you feel as hopeless as I did at the beginning, give it a go! Read/watch a few Sparkfun/Adafruit tutorials! Learn about keyboard matrices, then fuck it up regardless… Then try again! It’s gonna be worth it in the end.


With all this very important (</yawn>) knowledge collected, it was time to design “the endgame”.

The Polygon phase

I started out with a little cardboard example to get a feel for a different thumb plane. Note that the main keywell layout was already identical to the final product – that was never in question.

It then grew into a working prototype which I’ve already posted about here.

I was generally satisfied with it, so I went ahead and coded a Python → OpenSCAD model, too, in preparation of a “real” build5. The last picture is an exploded view to help understand the 3D structure…

Here came a huge, work-related gap where I did not have the luxury to retrain myself with the new board/layout (or to think a whole lot about keyboards in general) as I constantly had to meet strict deadlines under insane pressure; with the good ol’ QWERTZ.

Transitioning to 2D

That “gap” (and the associated pressure, thankfully) ended this April, by which time I’ve reconsidered a few things due to a) the prototyping experience, b) the designs I kept seeing on occasional visits to reddit, and c) the notion and pursuit of minimalism that kept creeping into my life.

First, no more split; I “glued” the board together. I’m not saying that split is somehow worse in any way, as it might be optimal for people who frequently change up their board’s position. But for me, it always ended up in the same configuration, and it was just an additional hassle to align it every time. A one piece design also makes transportation easier and the whole board is much more “compact” as a result.

The other big change was abandoning the separate thumb plane and the tenting – so basically everything 3D. Again, I’m definitely not saying that 3D stuff is bad. But we’re well past objective, preparatory research now, and my subjective opinion was that while they weren’t “bad” or “uncomfortable”, they didn’t add much either. From an ergonomic POV, I found that the most important angle is the inward one, not the upward, anyway. The 3D stuff did, however, significantly take away from mobility and compactness (again), and also kind of killed the whole aesthetic of the board for me.

So I decided to redesign the whole thing strictly on the 2D plane. Of course when I say “the whole thing”, I really mean the thumb area, which I modeled as an actual arc with a given thumb radius. This, combined with the aforementioned 1DFH principle, lead to a thumb fan that reached well below the usual amount… But hey, if that’s where my fingers are, then so be it.

The “redesign” happened with the use of Maker.js as I (unsurprisingly) still didn’t want to draw if I could avoid it. In the spirit of quick prototyping, I focused solely on the positioning of the keys and practically everything else was a simple placeholder. For example, the whole outline was just an automatically generated outline of the union of the keyholes. It had jagged edges, and it looked very “industrial”, but that was good enough. As I said in (the first half of) a comment: “I’m totally unapologetic about the looks as the design is completely driven by ergonomics decisions…”.

When it reached a state where it could be unironically called “beta”, I had a steel keyplate cut, desoldered the Polygon switches for repurposing, manually wired in a Pro Micro (that I attached to the bottom of the plate using blu tack), put a piece of cardboard underneath it all (that I cut out from a fucking cucumber box I scavenged from a nearby shop), and called it “done”. I posted about the rough first draft here.

Aww, isn’t that cute? I even had a little hatch on the bottom (for accessing the reset button) that opened/closed and was held by a piece of tape…

Going wireless, and fancy

As soon as I started using the prototype – even despite the frustration that came from the incredible slowdown I was experiencing – I knew that this layout was the shit6! So, naturally, I immediately started preparations for a “real” build… This time, for sure!

The first thing I had to consider was that wireless operation required me to say goodbye to my beloved Pro Micros. Let me seize this opportunity to thank SouthPawEngineer (pentagram in ink with a 65g Zealio switch in the middle) and iamjoric for their help with the wireless aspects, and just generally orienting me to have an idea about what I should consider. After checking my options from a shitton of perspectives (availability, reliability, replaceability, whatever-ability… and price), I chose the Adafruit Feather nRF52 Bluefruit LE as the controller. It was really an “all-in-one” package for me, as it ticked every box I cared about:

  • high quality
  • native Bluetooth LE
  • built-in LiPo charging
  • USB programmability with auto reset
  • plenty of GPIOs

It was, however, larger than the Pro Micro, and (of course) also required a battery, so I had to make room. As the middle region couldn’t accommodate both side-by-side, I initially went for a thicker construction where the battery would go under one of the keywells – more on this later.

The other thing I had to address was the “polish” of the outlines, the cover layers, and the fancy-ization in general. And here is where the second half of the above-mentioned comment comes into play: “I’m totally unapologetic about the looks as the design is completely driven by ergonomics decisions… The only minor things I have my ‘aesthetic influence’ in is how the corners are rounded, how thick the margin is, how big the back opening is, etc.”. Suffice it to say, that “minor influence” quickly turned into an incredible OCD phase where I fought with gaps, joints, and corner roundings so minor that you’d barely even notice. The amount of time I pissed away trying to make them look juuuust right was nothing short of ridiculous. I, of course, also scheduled a complete rewrite while I was at it so the code can be as logically structured and “elegant” as I can make it (because why wouldn’t I?).

The results of this phase were summarized in this quality shitpost, full of comparisons to butterflies, briefs, and even uteruses.

A visit from the “Low Profile Police”

This is the point in the process where I thought I was done with the design, but then DanL4 came along to fuck everything up! He suggested that I somehow solve the positioning of the controller and the battery above/below each other to cram everything in the middle, thereby making the whole build about 4 mm (1 wood layer) thinner. (Always with the making it thinner spiel… I’m surprised I was even allowed to use full size MX switches at all!) The shameful part here is that I’d actually thought about this before and concluded that they just couldn’t fit. Even though I really wanted them to, as it would have also solved the problem of having to mount the battery to the bottom, while everything else was mounted to the top part (thereby making disassembly much less safe/convenient).

That is, until Dan suggested that since the top cover is 2 wood layers anyway, why don’t we just hollow out the lower one to house the controller, so the battery (that would go below it, but still connected to the top assembly somehow) could now fit in the same space the switches take up anyway. I did my measurements and then (both happily and begrudgingly) conceded that he was right. And since it lead to an objectively better design, I couldn’t bury my head in the sand… so I started to redesign stuff.

The mounting of the controller changed from threaded inserts (that I also planned for the side screws) to manually glued nuts inside the hollow space. And if we take the 4 mm hollow + the 1.5 mm keyplate into account, I only needed another 1.5 mm of rise to be able to accommodate the battery as well, which I solved with a dedicated “battery plate” that then could be screwed to the top part. Take a look at some of the interesting-looking outlines born this way (undercover7, keyplate, dampener, batteryplate, and middle_top, respectively):

Dan continues to be an impromptu product manager on the project – spoiler alert: he’s a product manager. He pokes holes in every assumption I make, and having to explain myself to someone who’s also very knowledgable in this field makes the end result objectively better. He’s definitely the person who’s helped me the most during this whole madness. Thanks a lot, man!


The name comes from the Tim Burton version of Alice in Wonderland, where they actually named the Caterpillar. And connecting this keyboard project to the Caterpillar from Alice is pertinent for multiple reasons:

  • the keyboard looks like a butterfly…
  • my all-time computer (and the in-progress environment I’m cultivating on it) is called Alice, and it’s very fitting to have the “Absolem” blow letters to her over the air:
  • in the Tim Burton version (again), Absolem helped Alice & co. interpret the Oracle, which would make me the Oracle in this analogy, and that’s an ego-boost I can’t pass up!
  • plus it also sounds really cool, even if I do say so myself.

Logo stealing design

I knew that I wanted something butterfly-based, and also something geometric. What I also knew was that I suck at graphic design. So I started Googling “butterfly logo” and similarly creative terms hoping for the best. I even took side-quests like “hookah logo” or “caterpillar logo” but there was no clear winner. Until, that is, I tried to be as direct as possible and typed in “absolem logo”. It immediately lead me to absolem.pro, which – I assume – is a Russian hookah lounge franchise… with the perfect logo foundation! See if you can find the similarity:

So yeah, I totally stole borrowed inspiration from this, and did a few minor modifications:

  • the whole thing – like the keyboard plans themselves – is generated procedurally, so there are definitely some differences in the angles and the distances
  • the tail is now short, because why would it be long in the first place?
  • it now has a “head” with little “antennae”
  • it has 5 “bars” on either side (representing the 5 columns of the board), of which only 3 remains for the lower region (representing the thumb keys)

The end result got engraved onto the cover layer, and I’m very pleased with how it turned out.


Ooookay, enough small talk, let’s get our hands dirty!


With the digital side of the design ready, it is time to discuss the specifics of building an actual, physical thing. The general guidelines I followed can be summed up as:

  • natural – it should feel more like an instrument than a cheap peripheral
  • high profile – anything that is “slim” is rarely “robust”
  • minimalistic – so a clean, no frills look, no wires, and definitely NO visible screws
  • quiet – only the bare minimum noise that’s required for ergonomic operation
  • heirloom-grade3 – a construction quality that would actually make it feasible to leave one of these to the next generation.

Translating the above to materials led to:

  • 4 mm thick oak layers for the cover/undercover and the “middles”,
  • 1.5 and 1 mm thick stainless steel layers for the switch plate and the bottom cover, respectively,
  • 3 mm thick embossed neoprene for the anti-slip bottom and the internal sound dampening (it didn’t have to be embossed, but it looked so cool under my mousepad that I couldn’t resist), and
  • blank PBT keycaps for comfort, and a minimal, clean look (and to force me to actually learn my fucking keymap).


Getting all these required a lot of calls and messages to a lot of different people, and that doesn’t even include having them cut according to the plans. Oak boards of the desired thickness are only sold by one webshop in the whole country. For the neoprene, I had to make a specific deal with a Chinese wholesale manufacturer to send me A4 sized “samples” (and where the shipping cost was a whole magnitude larger than the material). Regarding lasercutting, a lot of companies don’t take small, one-off orders so I had to shop around quite a bit to find two that accepted the metalwork and the wood/neoprene, respectively.

As for the other “ingredients”, the Zilents came straight from the official store in Canada, and all the rest came from Aliexpress:

For the assembly, I also needed:

  • sandpaper in different grits
  • alcohol and Q-tips to clean the plates and the neoprene sides
  • universal glue and a professional, dedicated “spreading toothpick” – to keep the threaded inserts in place, and also to glue the wooden layer-pairs together (I would use dedicated wood glue next time)
  • a hot glue gun – to affix the controller nuts and to glue the switches in place
  • clamps (I had 4, but I would probably buy more next time)
  • wood stain, rags, brushes
  • a soldering kit, and a good wire stripper for the wiring
  • a 1.5 mm allen key for the screws

Note that while the above shopping list looks quite nice and tidy, it’s the result of probably half a dozen (or more) trips to different stores on as many occasions when I just remembered that “Oh, I need that as well!”. And then a looot of waiting for everything to arrive.

As for how much all this cost me, I can really only guess because of all the tools I didn’t have and all the experimentation I did – what we include here matters a lot. In any case, I’d say I’ve spent somewhere in the neighborhood of $400 (so far!). If it ever came to making this “repeatable”, though, with a little streamlining, and we wouldn’t include the switches and the keycaps (as most estimates usually don’t), it could be well below $200.

Two short, related anecdotes:

  1. The metal lasercutting shop is in an industrial park a few kilometers out of town, and when we arrived there to pick up the plates with my wife by bike, some workers were laughing at us pretty hard. They said that they have been doing this for a while, and have encountered the whole spectrum of cars, vans, and trucks at the shop, but nobody has ever come to collect their order by bike yet… :)

  2. When asking around about wood processing options in a local shop, I brought along the current prototype for illustration. The 2 guys there were really helpful (they referred me to the lasercutter I ended up with), but there was also a woman who acted really condescendingly towards me the whole time, and whose only contribution to the conversation was a careless “That’s not possible… You have to forget about this!”. I’d just like to state for the record that it was indeed possible, and send a heartfelt “Fuck you!” to that “lady” through this channel as well…


Even most of the assembly consisted of waiting with all those drying and glue-setting times. It also didn’t help much that I’m a complete n00b, so I would rather not tell you the actual (ungodly amount of) time it took me to do this.

Let’s start with the materials like wood (raw & cut), which I’ve sanded both before and after the cut.

Metal (raw from the cutter’s & brushed). I first cleaned the stains with rubbing alcohol and then brushed them with a metal sanding paper to make them look nice. Who’d’ve thought that you can make brushed metal by just taking a piece of metal, and… you know… brushing it?! My deep DIY experience shines here, but this really surprised me; I expected something more complicated.

And the embossed neoprene (raw & cut). The lasercutting left some staining residue on the sides, so these pieces required an additional round with some water and several Q-tips to properly clean.

Here’s the first, tentative assembly showing how the part will look like together.

For a “feel test”, I put the switches in and some (scavenged) caps on. As you can see from the left pic, the pinky and the ring fingers are deemed the weakest with 62g springs. Then comes the middle finger, which is – well – the middle, with 65g. The index finger is stronger, so it gets 67g springs – except for when it has to do diagonal movement, in which case we lighten its load to be the same as for the middle. And finally the thumbs get the heaviest switches with 78g springs. I have to say the feel(ios) are excellent! There’s definitely going to be an adjustment period as it feels a little heavier than pure browns overall, but it’s gonna be worth it.

So I started the actual DIY with testing the (dark walnut) stain on one of the leftover wood pieces, and rounding down the edges on the cover where the thumbs might be hindered later on.

Then came the installation of the threaded inserts into the undercover layer, and the gluing together of the two middle parts. During these steps I managed to break the wood not once, but twice – one for each step. Fortunately, I was just in the middle of gluing wood, so I glued it back together, and that was it. No one expected this to be a factory level precision assembly anyway, and these little scars and imperfections just give the finished piece some character…

The gluing of the two middle layers was easier as they had holes to begin with. But the gluing of the cover (which is deliberately screw-free) and the undercover needed some funny-looking clamping.

I used one of the spare wood boards to clamp flat the non-slip neoprene layer while its glue set. And on the right, we can see two of the well-known cucumber boxes (which also gave the bottom of the prototype) covering the wood pieces while the first round of stain dries.

I don’t know the exact point in the process where the wood got a little warped, but it did. So I applied a little heat (I literally ironed the layers) and then clamped them to a straight surface to cool “the right way”. Anyways, the results are looking good so far!

I then applied a second coat of the stain, which turned out wonderfully!

Next, I started progress on the keyplate by gluing on the dampener neoprene. I also put in the switches again in preparation of the wiring, when I noticed that the switches fit veeery loosely in the holes, and I won’t be able to use a keycap puller without removing the switch as well…

… so I glued the fuckers in with some (that is, a lot of) hot glue. Now they were ready for the keycaps that also arrived in the meantime!

Gluing the nuts to make a sort of “controller socket” went about as lamely as it could. It wasn’t enough that I could barely control the glue gun and made a mess (hey, it’s inside, no one’s gonna see it…), but it also turned out that the Bluefruit couldn’t accept one of the mounting screws because it would have physically interfered with some on-board components. So I made it three-legged. At least it will never wobble, right? The point here is that improvisation is necessary, no matter how much preparation you do in advance.

I also taped (and then trimmed) the backplate to protect against accidental shorts.

And (in a very long-story-short way) the “shell” of the Absolem was done – even though it was still completely empty inside…


To have the monster come alive, I first had to do some prep work (again). Namely, bend diodes, measure and “loop-ize” wire, and select the correct screw lengths, among others.

Then came the time to solder all the things! (First the diodes, then the rows, and lastly the columns.)

The soldering continues on the controller side as well with the columns on the top, and the two sets of rows on the bottom.

It all just looks so nice and orderly this way that I can almost believe that it’s not the same build where I screwed most things up, and made a mess even when I didn’t… To make things feel a bit more realistic, let me just show you that I used tack to hold wires in place while soldering (left) and that I also managed to singe the battery connector with the soldering iron (right).

Fitting and connecting the controller went relatively smoothly, but then came the battery, which proved a little more difficult. You see, if I’d paid attention to the Bluefruit wiring diagram I might have noticed which side expects the red wire and which the black. But I didn’t even pay attention to this part as I thought that everyone’s following the same standards.

Well, no. I had the misfortune of ordering JST connectors with the exact opposite wiring, which I only realized after a kind soul helped me out on the Adafruit forums. This is why you always go for batteries with the protection circuitry built-in! So I reversed the polarity, taped the battery to its plate, and tada…

As proof of life, you can see the yellow light on the left – showing that the battery is charging – and the red light on the right – which I made blink from a test sketch… Victory!

The finished product

This whole process has often reminded me of jedi knights building their own lightsabers… And finally I was ready to “graduate”, as the build was finished!

Well, physically… It was nothing more at this stage than a fancy desk ornament as it did exactly nothing, – except blink a little led – but it was finished nevertheless! You’ll have to forgive me if I sound a little too fucking proud… Anyways, feast your eyes8!


Okay, so how do we make the shiny new build actually functional? QMK was fine in the prototyping phase, of course, but now we’re dealing with a Bluefruit… Well, I looked into the alternatives, but didn’t find an “as-is” solution (just like with the physical keyboard itself).

The closest I got was jpconstantineau’s firmware (who I’d also like to thank here, his work really made mine a lot easier). And although the BlueMicro_BLE is a good starting point, it didn’t support everything I needed from my QMK days.

While I was looking into it (and QMK, and others), though, I caught the bug again. Why don’t I write my own firmware, too? I could do better than the current state of the art! Better being a more “elegant”, object-oriented logical design, I mean. I know too little about the low-level stuff to even think about competing in pure performance… But modeling a generic keyboard firmware infrastructure was right up my alley. And it also promised to be a good chance to grow as a programmer, and to be in complete control of both the hardware and the software aspects of the Absolem. So this is exactly what I did.

I don’t want to go into too many details here – this post is almost a book already – so I’ll probably write about the firmware separately in another post if there’s any interest. In the meantime, you can check out the code here.

The important thing is that as of about a week ago, I was able to almost completely (99%) replicate my old QMK config. And with such grace that I just couldn’t stop patting myself on the back. :D Seriously, though, I think it turned out pretty okay… There’s still a lot to do on this front, but I’m covered for now.


My keymap is available here, which (similarly to the physical aspects of the builds) was very much born in the spirit of “go hard or go home”. The alpha layer is Colemak Mod-DH without the punctuation, and the thumbfan is mostly the “destress the pinky” zone leaning heavily on mod/tap and layers. The numbers are laid out like on a numpad, and the navigation layer (with full modifier support on the left) is designed with text editing in mind. There are ergonomic consideration even on the symbol layer, where I placed the symbols to more/less comfortable positions according to their relative frequency in both natural language corpora (corpuses?) and programming language source code.

Adjustment progress

As it’s an enormous shift from full size QWERTZ to 36 key ergo Colemak, I’ve decided to put in the hours and learn proper touch typing this time. It was pretty disheartening to go from around 60 wpm to below 10 in a blink. But I then completed the typing.com tutorial and reached ~30wpm, which provided enough familiarity to start the slow but steady climb over on NitroType. After daily practice for over 4 months (which I log here), I’m currently at an average of about 80 wpm with occasional visits in the 90s… Not part of the elite (yet) but progress nevertheless.

I also frequently use Keybr.com (like, a LOT), 10fastfingers, typeracer, and SpeedCoder for some varied practice.


I then decided that instead of writing a simple, short, and sweet reddit post about this, I’d write a more long-winded blog thingy to serve as the canonical home for the Absolem project. In it, I divide the whole process into individual sections and talk about each topic in detail. There’s even a section in it which is about this writeup, in which I discuss why I wrote it, and how it consists of sections, one of which is about this exact writeup, in which… Whoops, infinite loop! break;

Future work

Okay, if you want to nitpick, I did really say “endgame” in this post several times, which should mean that there’s no “future work” remaining…

In my defense, though, I’d like to say that:

  1. my TODO list is just about refinement, so we can justifiably say that it’s v1.0 already;
  2. I could have waited with this post until all the refinements are done as well… but just couldn’t didn’t want to; and
  3. I’ve always planned to have 2 boards anyway (so that I have a backup if the main one ever needs service) in which case it would be foolish of me not to learn from the first round’s experience and correct any minor imperfections.

So, even though I’m re(eeee)aly pleased with how things turned out, here are some further improvement ideas I have in mind:

  • Physical:
    • I want a PCB – despite the fact that I clearly stated in the past that I don’t want a PCB. I mean, I still stand by what I said there about the potential durability and fixability differences, but what I didn’t consider with enough weight then is that it could hold both the Feather and the battery in place and I wouldn’t need to glue extra nuts to the underside of the cover layer (I hate gluing…). It would also add more stability to the switches so I wouldn’t need to glue those in either (Have I mentioned yet that I hate gluing?). Plus it would look much more “professional” and clean. And PCBs are not that expensive… And it would make it much easier for other people to potentially build their own. So yeah, I want a PCB. If you’re a PCB wizard and want to help, I’d love to hear from you!
    • I want a thumb valley – as the topmost cover layer (while beautiful as it is) still bothers my thumbs a little at certain angles, even after I deliberately filed away most of the edges there in anticipation of this exact thing… Again, it’s not prohibitive or anything; it’s barely noticeable most of the time. But if I were to do this again (which I will), I would probably leave that region completely out as early as the lasercutting stage and then merge it together with the undercover layer to make a more pronounced “valley”. I’m probably not describing this very visually, but you’ll see the difference when I get to it…
    • I want a sleeve – to replace my trusty keyboard carrying travel towelTM. It’s not the functionality, it’s the prestige of a dedicated (and, of course, branded) carry case. Don’t know much about how I want to solve this one yet, so it can end up a sleeve, or a hard case, or some sort of origami-ish furoshiki cloth… Time will tell; stay tuned…
  • Firmware-related:
    • I intend to solve various energy efficiency issues down the line, including a “deep” sleep, plus also a kind of middle ground between full activity and full sleep that would kick in earlier and where the matrix scanning is slower.
    • Yet another kind of sleep, this time for the Bluetooth connection polling, which could then be restarted of course, it just wouldn’t waste the battery in an infinite loop.
    • Measuring (and then hopefully further optimizing) battery usage.
    • Finish the initial ideas sprinkled throughout the repo regarding general use and personalization, and add a build script that streamlines the process.
    • Separate the firmware part to its own repository (and add that as a submodule of the original one) to make it more accessible for potential other projects not associated with the physical aspects of the Absolem.

I’ll probably write another post (and link it here) when I get around to addressing all these.


I have a controlled, but ever present tendency to “fanboy” over something if it really deserves it. For example, I’m a proud owner of a Secrid Miniwallet and I consider it a cool “club” to be a part of. And I would very much like it if there was an official “Absolem club”, even if it’s just the two of us!

So all I ask of you, dear reader, is that if you ever decide that the Absolem is a good match for you and you build one, please contact me so that I can add you to the club! Or you could contact me to build you one; then I’ll know to add you automatically! It’ll be great, I promise… We could even wear a badge! Too much? Okay, I’ll stop now.

Anyways, there’ll be a list of some sort here, if there’s ever someone other than me… :)

Update: and there is someone. Wild Absolem builds have been spotted in Israel, the Netherlands, and France! Plus there are “derivatives” as well where the Absolem was credited as an influence. A real club list is in order now, but until then there’s an official Absolem Club Discord, so come join the discussion!

That’s it, folks!

If you’ve seriously read all the way down here, then you’re awesome! I hope I managed to keep it interesting. Now go and upvote the corresponding reddit post and star my GitHub repo to earn an official, non-refundable, virtual high-five!

Contact: you can find me by mail, on GitHub, on Reddit, or on Discord, and if you have any feedback regarding the Absolem, I’d love to hear from you!


  • 2019.09.07. – Original publication.
  • 2019.09.07. – I already stand corrected: what I referenced as “ergowrap” is actually the ErgoWarp. My apologies!
  • 2020.06.25. – Fixed the firmware link to point to its new home, plus tried to add a favicon.
  • 2020.06.28. – Actually added a favicon… Plus some theme tweaks and some club-related updates.

Acknowledgment: Besides the people already mentioned it the post, I’d like to thank my wife, who put up with me during this year-long process, and whose first reactions weren’t “You’re doing WHAT?!” and “It costs HOW MUCH?!” :)

Disclaimer: This post contains lots of pictures and references that are obviously not my work, and I’m not claiming that they are. GIFs and jokes come from quick Google searches, and I’m not going to put a “source: something…” under each one. Use your best judgement.


1: notice how my nick is “zealot”, which is just a cooler name for fanatic. (I reject the religious connotation, though.)

2: “best” is arguable here, I admit. But, despite my every effort to keep this as objective as possible, the whole topic is somewhat subjective, so bear with me.

3: “heirloom-grade” was a phrase I first encountered on the KeyboardIO site, but subconsciously it has been the driving force behind a lot of the design and material choices I’d made even before then. I freely use it, though, as I find it sums up my intentions really well.

4: My crafts teacher in elementary school (may he rest in peace) once said to me that if there’s ever anything to fix around the house, I should just call someone over :D

5: If you think that it’s a shame that I didn’t go in this 3D-ish direction, then feel free to go dumpster-diving into the repo’s history, and fish out the model to use/modify.

6: I just love how when something’s shit, that’s bad. But when something’s the shit, it’s really good. Also relevant is the crazy relationship between horrible/horrific/terrible and terrific. English is (the?) shit. :P

7: this layer is not called “undercover” because it did some incognito work for the FBI. It’s just literally under the cover layer.

8: As you might have noticed from the potato-quality photos, I’m not a photographer. So please do both of us a favour and concentrate on the subject matter of the pics, not the quality!