rawdata.stream - Home

The August (optimized Dvorak) keyboard layout

Unknown — Thu, 31 Jul 2025 00:00:00 +0000

Twenty years ago (2005), I learned to touch type. Before that, I was a "pretty fast" hunt-and-peck typist (I recall testing myself at 40-something words per minute). I had just finished grad school, and decided it was long past time to learn to touch type. I got some kind of typing tutor program, and it offered me Qwerty or Dvorak as my options. I thought Dvorak was the obvious choice, so I chose Dvorak. If Colemak had been an option at that time, I might very well have chosen that, but it was released in 01-01-2006, according to its FAQ. Anyway, I really had no problems with Dvorak. I learned to type pretty fast with it, probably in the ballpark of 90 wpm (based on my memory of doing some speed tests). However, I eventually got the bug to create my own optimized layout. This seems to be a thing that people like me, who have something off-kilter in their head, become hell-bent on doing. I recall having been originally inspired by the Klausler layout, and his description of how he had created a custom layout.

My opinion, after having dove head first into the fish tank of alternate keyboard layouts, is that Dvorak is actually a surprisingly good layout when you consider that August Dvorak did not have access to all the modern technology that keyboard layout enthusiasts have (most notably optimization algorithms such as Simulated Annealing and the computers needed to run them). He created a layout so good that, in my opinion, most people who already know how to type Dvorak do not need to switch to anything else. However, if you yourself decide to dive into that fish tank, just like I did, you may be alerted to some of the flaws that Dvorak has. But Dvorak is a vast improvement over Qwerty, and that is an understatement. And Qwerty is so ubiquitous that, from a certain vantage point, it seems outrageous to point out any flaws in Dvorak. If you don't go down the rabbit hole yourself, you likely cannot conceive of how ridiculously inefficient the Qwerty layout is. It is truly a monstrosity, by almost any measure of typing efficiency. The very fact that it was designed so that you could type the word "typewriter" on the top row (not even the home row) is a testament to how braindead it is. And yet here we are. And what's amazing is that people have learned to type unbelievably fast with Qwerty keyboards. So for anyone who wants to defend its design, the only defense you have is that the human brain is so adaptable that even the Qwerty keyboard cannot defeat it.

About fifteen years ago, I considered switching to Colemak. Actually, I started to, but it was just too different from Dvorak. Not only was virtually every letter in a different place, but all the punctuation, too. And Colemak's whole premise was that it's an improved Qwerty-based layout: by changing the position of only a modest number of keys on the qwerty layout, Shai Coleman obtained a layout that's close-to-optimal. Of course, nobody can agree on what's optimal, but most people will agree that Colemak (like Dvorak) is vastly better than Qwerty. And since Colemak is easier for a Qwerty typist to switch to than Dvorak, Qwerty typists need not consider Dvorak in the modern era; just learn Colemak, unless you feel like going down the rabbit hole of evaluating other alternatives that might be marginally better in certain ways. Like me trying to switch from Dvorak to Colemak, a Qwerty typist trying to switch to Dvorak means essentially re-learning every key on the keyboard.

Ten years ago, I decided I wanted to play the game of creating an optimized layout of my own. I was already typing on an optimized layout, so why go through the trouble of re-learning every key on the keyboard by switching to Colemak just to fix a few minor quibbles? I was mainly motivated by the joy of optimization, not so much any actual desire to type faster than I could on Dvorak. I just wanted a fun project, and moving a few keys around on Dvorak to get a better layout sounded like fun to me. So off I went. And the next thing you know "a few keys" turned into "almost every letter". It's hard to stop once you get started modifying things. Some of the details of that project can be found in my Colemak forum post from ten years ago, but I would like to write something up about how I developed the typing effort model and optimization code here if I can find some time to do that.

I call my keyboard layout "August", which is Dvorak's first name, and also the month I was born in. I did not change the punctuation keys in the upper right part of the keyboard from the way they are on the Dvorak layout. The final form that I have been using for the last 10 years is:

However, it's a modified version of the one that I would have chosen if I typed the way a proper typist types on an ANSI keyboard. When I learned to type on a Dvorak layout, I chose to hit certain keys on my left hand with the "wrong" fingers, because it was more comfortable. The variant of Colemak called Colemak Mod-DH is similarly inspired by the fact that the proper way to type is less comfortable for many people. The problem is the asymmetry of the standard ANSI keyboard (the keyboard pattern that is used in the United States), which is not left/right symmetric (stare at a keyboard for a moment and you'll see), so what feels natural on my right hand cannot be mirrored on my left hand. Anyway, my original optimized layout for typists who use the "correct" fingerings is:

I'm not recommending anyone else use this, but that's what I consider the "official" August layout, even if I changed things up a bit to adapt to my oddball fingerings. It's been ten years since I began using the August layout, and I have no complaints. I just think about the words I want to type and my fingers move in the sequence they need to to make those words appear on the screen, just like anyone else that types. I type about 100 wpm, according to the typing tests I occasionally test myself with. That's pretty fast, but it's not going to win any competitions. You really can learn a new keyboard layout fairly quickly once you decide to switch. If you already type 100 wpm or more on a Qwerty layout, then I am not sure you'll gain a whole lot of speed by switching to an optimized layout, but you are likely to improve the comfort of typing. But if you type on a Dvorak layout, there's really not much reason to switch. Some people complain about the "L" and "S" keys on Dvorak both being on the pinky, and I suppose I was one of those people, but it's not likely to be worth learning a whole new layout because of it. But that's what I did.

Here is a web page that allows you to compare different layouts, and shows some metrics for how they perform for things like finger distance traveled. I generated the layout image using the Colemak Mod-DH analysis tool. The github repository for the layout is here.

FORTRAN 77 to Modern Fortran conversion script

Unknown — Wed, 30 Jul 2025 00:00:00 +0000

Some time ago, I had posted to bitbucket my code for translating FORTRAN 77 into modern Fortran (Fortran >= 90). I guess at some point bitbucket deleted the repository. So I'm reposting it to github. f2f was originally written in about 2003, when I was in grad school, writing Fortran code for my thesis. I was too young to appreciate the syntax of Fortran 77, and too stubborn to just accept my fate. So I wanted to convert all the Fortran 77 code I had to use into Fortran 90. That's what prompted me to write this script, and I learned Perl to do it. At the time, Perl seemed to be the obvious language to use for this purpose (and might still be), and yet the other attempts to make conversion scripts were written in Fortran, if I recall. I think Fortran is a terrible language to use for translating a long chunk of text into a different chunk of text. Some years later, Bob Apthorpe rewrote my Perl code and sent it to me. Given that f2f was the first Perl script I had ever written, and his version just looked better, I figured it was probably better, and I just accepted his "pull request" without very much testing (pull request in quotes because git did not even exist then). Even so, since it was essentially a complete rewrite, and since the version control at the time seemed particularly archaic and convoluted to me, I just kept them both and made both available. And since I don't have any other history of the code, I'll just post both of them here for historical purposes. Most likely you won't need the older version, but if one version mangles your code, then maybe give the other a try. But given that it was my first Perl script, it has a special place in my heart. There are no formal tests, and there are so many edge cases in Fortran 77 that it very well might not produce correct output for your very specific code. I've never actually seen it produce run-time bugs, only fairly obvious syntax errors that the compiler catches. But it has always been useful to me, when I needed it, and to some of you, too. So here it is.

Serine for brain and retinal health

Unknown — Sun, 13 Oct 2024 00:00:00 +0000

This is some notes I made when I was researching macular degenaration, mostly just a curated list of research papers that I wanted to summarize. I am NOT a biochemist, I just love reading about it, and trying to understand it, so do not take this as anything more than a non-expert trying to make sense of it.

Summary

I'm going to make a case for why I think the amino acid L-serine would benefit brain and retinal health, especially as we get older. Cells that must sustain high rates of glucose consumption to function should be the first ones to deteriorate when insulin resistance or other age-related problems reduce the rate at which they can process glucose, maybe because housekeeping functions of the cell need some of the carbons from the incoming stream of glucose in order to synthesize important molecules that deal with the mess created by the cell's high metabolism. For some of those important molecules, like glycine and glutathione, the first step is to divert some of that glucose toward serine synthesis. In the brain and the retina, it's really a network of inter-dependent cells that becomes dysfunctional, but they still rely on a steady stream of glucose, and providing an alternative (dietary) source of serine may help alleviate some of the stress and prevent damage to the cells.

Introduction

Serine is a "non-essential" amino acid, which is a funny way of saying it's so essential that you can't rely on your diet to provide enough to cover what your cells need to survive. It's not essential to get in your diet, because your cells have the ability to synthesize it from scratch (using glucose). There are many different kinds of cells in your body, and some of them have a particularly high demand for glucose; they are the Formula 1 race cars of the cellular world. Two kinds of cells that fall into that category are astrocytes and retinal cells. Some glucose is refined into ATP and used as an energy source for the cell, and some is used to build other important molecules for cellular growth and operational needs. Serine is one of those important molecules, and unlike many of the others, it can be obtained from your diet, which takes that burden off of your cells and allows the glucose to be used for other critical things. For most cells, diverting some glucose into serine synthesis is not a big deal; but for the Formula 1 race cars of the cellular world, it might be the difference between winning the race and crashing-and-burning.

[I'm trying to imagine what it looks like when your astrocytes can't metabolize glucose as fast as they did when you were younger. I suspect that your cells start out with the ability to metabolize glucose faster than they really need to, and then that ability deteriorates (with age or other reasons) until it becomes the rate limiting step for glycolytic flux. After that, is it like underclocking a CPU, does your brain just slow down? It's probably not quite that simple, but I suspect that is one of the effects (see Decreased Alpha Peak Frequency Is Linked to Episodic Memory Impairment in Pathological Aging).]

Motivation

I enjoy biochemistry and neuroscience, and I first got interested in this topic when I was reading about Alzheimer's and glycolysis. But my mom has macular degeneration, and is going through treatments for that, and there seems to be a lot of overlap between some of the metabolic features of Alzheimer's and macular degeneration, even if they don't necessarily happen to the same people.

Research Papers

L-Serine: a Naturally-Occurring Amino Acid with Therapeutic Potential

L-Serine, an Endogenous Amino Acid, Is a Potential Neuroprotective Agent for Neurological Disease and Injury

Glycine and aging: Evidence and mechanisms

Macular Degeneration and Retinal Health

A Metabolic Landscape for Maintaining Retina Integrity and Function. Viegas and Neuhauss, 2021

This is a great overview of the retina from a metabolic perspective, and the figures are fantastic! It barely discusses serine, but Figure 1 helps illustrate the opportunity cost (in cells where glycolysis is a choke point) of using glucose to synthesize serine instead of pyruvate.

Serine biosynthesis defect due to haploinsufficiency of PHGDH causes retinal disease (Eade et al., 2021)

A genetic inability to synthesize sufficient serine will cause a kind degeneration of the macula called macular telangiectasia type 2 (MacTel) that is different from the more typical presentation of "macular degeneration"

Serine and glycine physiology reversibly modulate retinal and peripheral nerve function (Lim et al., 2024)

Highlights the potential importance of serine supplementation for treating MacTel
The best excerpt, from the Discussion: We demonstrate that circulating serine is the primary supplier of serine to the eye, highlighting the importance of systemic sources (e.g., liver, kidney, and diet). De novo synthesis from glucose by PHGDH and, to a lesser extent, glycine via the glycine cleavage system and one-carbon metabolism also contribute. Unlike most tissues, the eye appears to have activity of all available pathways to obtain serine. The liver and kidney primarily rely on glycine transport and one-carbon metabolism, rather than producing serine from glucose, to ensure stable circulating levels of these amino acids. By contrast, the brain heavily depends on de novo serine biosynthesis in astrocytes. The retina, by heavily relying on circulating serine, appears to be highly susceptible to the function of peripheral tissues that impact serine levels (e.g., SGOC metabolism in the kidney and liver). This feature also offers an avenue by which supplying serine systemically can treat the retina. The uniquely diverse mechanisms by which the retina obtains serine likely reﬂect the importance of serine for retinal function and may explain why a systemic reduction of serine (as observed in MacTel) manifests in the retina.

Human macular Müller cells rely more on serine biosynthesis to combat oxidative stress than those from the periphery (Zhang et al., 2019)

The macular (central) retinal cells (specifically Müller glia cells) synthesize significantly more serine than cells of the peripheral retina. Of course, their metabolism is higher across the board (glycolytic flux, mitochondrial respiration, ATP production), but the authors also state that glucose is preferentially shifted toward de novo serine synthesis instead of glycolysis/TCA cycle, suggesting that serine may be especially important in the macula.

The Intersection of Serine Metabolism and Cellular Dysfunction in Retinal Degeneration (Sinha et al., 2020)

This paper reviews the ways in which retinal cells rely on serine
"perhaps the most vital part of serine metabolism is free radical scavenging in the entire retina via serine-derived scavengers like glycine and glutathione. It is hard to imagine that a single tissue could have such a broad and extensive dependency on serine homeostasis. Any dysregulation in serine mechanisms can result in a wide spectrum of retinopathies."

Alzheimer's Disease

Impairment of Glycolysis-Derived L-Serine Production in Astrocytes Contributes to Cognitive Deficits in Alzheimer’s Disease (Le Douce et al., 2020)

"Here, we show that the astrocytic L-serine biosynthesis pathway, which branches from glycolysis, is impaired in young AD mice and in AD patients."
"Supplementation with L-serine in the diet prevents both synaptic and behavioral deficits in AD mice. Our findings reveal that astrocytic glycolysis controls cognitive functions and suggest oral L-serine as a ready-to-use therapy for AD."
But see also PHGDH expression increases with progression of Alzheimer’s disease pathology and symptoms and then Glycolysis-derived L-serine levels versus PHGDH expression in Alzheimer’s disease for some followup discussion on the above study.

Glycolytic Metabolism, Brain Resilience, and Alzheimer’s Disease (Zhang et al., 2021)

Reviews research on the relationship between glycolysis and Alzheimer's disease, and makes it clear that glycolysis is central to the problems associated with Alzheimer's
"Accumulation of neurotoxic Aβ plaques and hyperphosphorylation of tau have long been considered the pathological hallmarks that contribute to synaptic disruption and neuronal loss in the brains of AD patients. It is reported that Aβ distributes variably among brain regions, with more deposition found in areas of high dependence on glycolysis... the regions with increased glycolysis in the resting state of healthy young adults closely mirror the later regional pattern where Aβ accumulates in the brains of AD patients"

GLP-1 improves the supportive ability of astrocytes to neurons by promoting aerobic glycolysis in Alzheimer's disease

Studied the effect of liraglutide (diabetes medication) on mouse cognition and astrocyte metabolism
GLP-1 mediated a shift from oxidative phosphorylation to glycolysis in the Alzheimer's animal and cellular models, and improved the astrocyte cells’ ability to support neurons as shown by increased cellular viability and robust dendrite and axon growth.

Iron deficiency

Unknown — Mon, 02 Sep 2024 00:00:00 +0000

One of the biggest themes of my life lately has been my iron status. In march, about six months ago, I had a routine blood test that showed I was anemic. My hemoglobin was 12.7 g/dL (not dangerously low, but outside the reference range for an adult male). My first thought was that it must be iron deficiency, because a) that's what most people assume, and b) I had my ferritin measured a few years ago (late 2021), and it was low. Ferritin is used as a measure of your "iron stores", since it is correlated with other measures of iron status in your body, so I already knew I was at risk of iron deficiency. What I didn't realize a few years ago is that low iron is not something that you fix with a few days or weeks of iron supplementation. It takes months or longer to raise your iron stores from depleted-to-the-point-of-anemia into the normal range. And the anemia itself is a separate but related problem; it also takes months for your red blood cells and hemoglobin numbers to fully recover, because the life cycle of red blood cells is several months, and they are slowly regenerated with more hemoglobin in them over that timeframe.

Since March of this year, I have done a deep dive into the science of iron. I dive deep on things a lot, and I would like to make a habit of writing up some of what I learn. In that spirit, I intend to write some things about iron. Maybe some of this will be of interest to someone on the web who finds my post through a web search, but I also want to write it up so that it doesn't just live in my head until it's forgotten. Because I, myself, may want a refresher some years from now. My intention is to put together a few different posts. One is about sources and losses of iron, and how one becomes deficient. Another is about iron-containing enzymes in your body, of which there are many, and they are rarely discussed in the context of iron deficiency (hemoglobin seems to be the only thing anyone thinks about). A third subject is iron supplementation, the protocol I followed, and why. I may have other things to say, but those are the ones I'm thinking about right now. [Months later, I still have not followed up on this, but since I'm really only writing for myself at this point, the only person I'm disappointing is me. Still, I hope I can get back to it someday]

End of the road for AnandTech

Unknown — Fri, 30 Aug 2024 00:00:00 +0000

I'm sad to read that AnandTech has posted its final article. I respect the writers and editors for the depth of the articles they published. You felt smarter after finishing one of their pieces, because it was written to actually educate you about the topics it covered. And in fact, it makes me want to go out and find other websites with such in-depth articles and actually support them by subscribing. I never did that with AnandTech (did they even offer subscriptions?), so I'm part of the problem. I hate advertising and the world it has created, so I would rather pay publishers directly, and now it's too late to save AnandTech. So it's a reminder that I (we) need to support the publishers that deliver high quality content. Their final article and the About page discuss the problem of sensationalist reporting, and how they aspired to avoid that fate. I am one of the people who wishes there were a good business model to promote the kind of writing that AnandTech published. Most of the comments on the Hacker News post about AnandTech's final article are also about this problem of sensationalist journalism, because I'm not the only one who feels a sense of loss, and a nostalgia for the days when AnandTech began.

Getting started

Unknown — Sun, 28 Jul 2024 00:00:00 +0000

I want to build the habit of writing, so I've created this website. I plan to talk more about why I want to build the habit of writing, but for now I'm just here to say that I'm going to, and that this is the starting point of that intention.

The website is called rawdata.stream because .stream domains are cheap ($6/year right now, I registered it through namecheap), and also because I want to encourage myself to write despite my own perfectionism, so I'm telling you up-front that it's going to be unrefined and all about content over form. In fact, "you" might not even exist, because I'm going to be writing for me, without any expectation that anyone wants to read this. These are the reasons it looks so plain and minimalistic visually. I may have more to say about the design (or lack thereof) later, too, because it's a purposeful choice to keep it as simple as possible, but I also just wanted to get started writing and not designing a website.

So for what it's worth, I'm using the Hugo static site generator, and the base theme I started with (and modified somewhat) is called smol. I could have used Wordpress or some other way to get started quickly, but I prefer a certain kind of simplicity that Wordpress just isn't, a kind of simplicity that to most people would seem less simple; to me it's about understanding the tools I'm using, having control over them, and not feeling locked in to somebody else's way of doing things.

So this is where I'm starting.