Epic Fail

Atul Gawande describes the Epic software system being rolled out in America’s hospitals.

It reads like a  potpourri of everything bad about enterprise IT. Standardize on endpoints instead of interoperability, Big Bang changes instead of incremental improvements, and failure to adhere to the simplest principles of usability.

The sad thing is that the litany of horrors in this article are all solved problems. Unfortunately, it seems that in the process of “professionalizing” usability, the discipline has lost its way.

Reading through the article, you can just tally up the violations of my proposed Usability Heuristics, and there’s very few issues described in the article that would not be eliminated by applying one of them.

The others would fall to simple principles like using battle-tested standards (ISO timestamps anyone?) and picking the right level of database normalization (it should be difficult or impossible to enter different variations of the same problem in “problem lists”, and easier to elaborate on existing problems).

There was a column of thirteen tabs on the left side of my screen, crowded with nearly identical terms: “chart review,” “results review,” “review flowsheet.”

I’m sure the tabs LOOKED nice, though. (Hint: maximize generality, minimize steps, progressive disclosure, viability.)

“Ordering a mammogram used to be one click,” she said. “Now I spend three extra clicks to put in a diagnosis. When I do a Pap smear, I have eleven clicks. It’s ‘Oh, who did it?’ Why not, by default, think that I did it?” She was almost shouting now. “I’m the one putting the order in. Why is it asking me what date, if the patient is in the office today? When do you think this actually happened? It is incredible!”

Sensible defaults can be helpful?! Who knew? (Hint: sensible defaults, minimize steps.)

This is probably my favorite (even though it’s not usability-related):

Last fall, the night before daylight-saving time ended, an all-user e-mail alert went out. The system did not have a way to record information when the hour from 1 a.m. to 1:59 a.m. repeated in the night. This was, for the system, a surprise event.

Face meet palm.

Date-and-time is a fundamental issue with all software and the layers of stupidity that must have signed off on a system that couldn’t cope with Daylight Savings boggles my mind.

A former colleague of mine linked to US Web Design System as if this were some kind of intrinsically Good Thing. Hilariously, the site itself does not appear to have been designed for accessibility or even decent semantic web, and blocks robots.

Even if the site itself were perfect, the bigger problems are that (a) there are plenty of similar open source projects, they could have just blessed one; (b) it’s a cosmetic standard, and (c) there’s pretty much no emphasis on the conceptual side of usability. So, at best it helps make government websites look nice and consistent.

(To be continued…)

Lost Productivity

A couple of months ago I listened to a Planet Money podcast discussing the mysterious slowdown in US productivity growth (the link is to one of several podcasts on this topic). Like most NPR content, the story got recycled through a number of different programs, such as Morning Edition.

The upshot was, that productivity — which is essentially GDP/work — has stalled since the — I dunno — 90s, and it doesn’t make sense given the apparent revolutions in technology — faster computers, better networks, etc.

Anyway, the upshot — and I’m basing this on memory because I can’t find the exact transcript — is that there’s a mysterious hole in productivity growth which, if it were filled, would add up to several trillion dollars worth of lost value added.

Well, I think it’s there to be found, because Free Open Source Software on its own adds up to several trillion dollars worth of stuff that hasn’t been measured by GDP.

Consider the dominant tech platforms of our time — Android and iOS. Both are fundamentally built on Open Source. If it weren’t for Open Source, iOS at minimum would have been significantly altered (let’s assume NeXTStep would have had a fully proprietary, but still fundamentally POSIX base) and Android could not have existed at all. Whatever was in their place would have had to pay trillions in licenses.

On a micro level, having worked through a series of tech booms from 1990 to the present — in the 90s, to do my job my employer or I had to spend about $2000-5000 for software licenses every year just to keep up-to-date with Photoshop, Director, Illustrator, Acrobat, Strata 3d, 3dsmax, Form*Z, and so on and so forth. By the mid aughts it was maybe $1000 per year and the software was So. Much. Better. Today, it’s probably down to less than $500.

And, in this same period, the number of people doing the kind of work I do, is done by far more people.

That’s just software. This phenomenon is also affecting hardware.

The big problem with this “lost productivity” is that the benefits are chiefly being reaped by billionaires.

Photo-backup, Hubic, and Other Stories

So, a year ago I started backing up to HubiC after my previous backup service decided to stop servicing retail customers. At the time, HubiC seemed on paper like a great option — potentially offering the simplicity and utility of DropBox with effectively unlimited storage capacity.

In practice, HubiC is useless. I’ve had a 2012 Mac Pro constantly connected to HubiC via a fast cable connection for 12 months and managed to back up only about 1/4 of the files I’ve pointed at it. The damn thing breaks down constantly. Every time I log in it wants me to change passwords. They just billed me for renewal, but I’m two days late and I can’t cancel my account without paying for another year. Good luck with that, guys.

Now, I haven’t been sitting on my hands while I watch HubiC fail to deliver on any of its promises. Most of my stuff continues to be backed up locally via Time Machine. The stuff I work on is stored in the cloud (iCloud, GitHub, Google Drive, and/or DropBox).

My big problem is photographs and video.

The disaster that is my Flickr account
Flickr’s auto-uploader turned my Flickr account into an unfixable mess

Now, when Flickr raised its “free” tier to offer 1TB of cloud storage for JPEGs, I jumped on that. It may not be RAW storage, but it’s better than nothing, and 1TB is enough space for a huge number of JPEGs. The big problem, Flickr’s (since discontinued) auto-uploader was so stupidly designed that it successfully rendered my Flickr account borderline useless (it created an album for every folder it found an image in, and it uploaded every image it found, including things like UI images inside applications and development trees, so I have “albums” comprising sprites from sample game development projects and logos for PHP templates) and Flickr’s account management tools look and work like something an intern abandoned in 2005, so just deleting stuff is an exercise in frustration. It looks to me like Flickr’s abandonware API isn’t really up to the task of even supporting a third party application to untangle the mess.

And of course, since Yahoo changed hands and various security scandals unfolded just logging into Yahoo accounts is a pain, and you need to navigate ads to even get into your account. Yahoo is the GeoCities of 2018.

Google photos in "action"
When I scroll to some arbitrary point on my timeline, this is what I see for some random (long) amount of time…

Recently, Google raised the “free” tier of photos.google.com to unlimited storage of photos where RAW files are JPEGs are processed into high-but-not-full-quality JPEGs on-the-fly. I’ve tried it and it’s pretty damn good. The uploader is smart enough to skip files that are clearly not important photos (e.g. too small, wrong format) and ignore obvious duplicates. The problems are (i) that the uploader application periodically just hangs and needs to be manually killed and restarted (ii) the web app seems to be weirdly slow and unreliable (I can log on with two machines side-by-side and they’ll see different subsets of my photos), (iii) no Apple TV support, and (iv) online photo editor seems to need one or two extra clicks to accomplish anything (but it’s a lot better than nothing). I’m pretty confident that my stuff is there, just not in my ability to see a given photo from a given machine on a given day. It’s certainly the most complete, easy-to-navigate, and shareable archive I’ve ever managed to create of my photographs. And if I can find a photo there, I can locate the original RAW image pretty easily.

Now, the absolute best system for dealing with my photographs thus far is iCloud. If I could simply rent 10TB from iCloud for a reasonable price (let’s say, $25/month) and get my Mac to automatically sync multiple volumes to iCloud, my problem would be solved. Obviously, I’m a happy Apple customer. If I were a more-than-casual Windows or Linux user then this would not be a useful option to me, and I’m not sure what I’d do, because I’m pretty sure there’s no equivalently seamless option for people who don’t want to pay the “Apple Tax”. Google Drive isn’t even a tolerable substitute for DropBox (although I think it has Sharepoint beaten).

Here’s where iCloud beats all other options:

  • I don’t need to think about it or do anything. (Well, on a desktop device, I need to NOT avoid storing my data in iCloud) If I take a photo, then it ends up in the cloud pretty quickly (basically, when the device gets recharged while on a LAN, if not sooner).
  • By default, full-resolution images are not propagated to all my devices (as would be the case for DropBox, or Hubic if it actually worked). Instead, as with everything in iCloud it’s available on-demand. (Indeed, it’s a bit reminiscent of the way iTunes deals with movies… superficially less convenient than pure streaming, but a lot more flexible and useful in practice).
  • If I ingest a RAW photo from a camera onto a device, then it’s in the cloud and available from any device on-demand (but it’s not wasting space on all my devices).
  • If I want to work on a photo, I can use the best native tools that are available on the device I’m using — seamlessly (although I’m inclined to actively avoid Adobe applications because Adobe’s workflow involves use of Adobe’s barely functional Cloud ecosystem).

The big problem — of course there has to be one — is that Apple’s highest storage tier is 2TB. I’m currently on 200GB which is plenty for the stuff I need that isn’t photos and videos, but hopelessly inadequate for photos and videos. 2TB (the next tier up, and it’s competitively priced) would be sufficient for my photos and videos if I were to curate them, but I don’t want to curate shit. I want to dump it in the cloud and not think about it.

Missing in Action

All of this adds up to a bunch of pretty disappointing non-solutions. Even though Apple provides a file sync system that works pretty well for personal photographs, it wouldn’t work for say a small photography business. (I guess you could use some kind of “family plan” but I’m pretty sure that would run you into weirdness pretty fast.) And it’s not like we’re talking advanced workflow support here — I just want my photos backed up and available.

Where is a tool that automatically detects blurred, underexposed, or overexposed photos and flags them as less worthy of backup? (Google’s photos app does a pretty good job of automatically correcting exposure, I wonder if it’s smart enough to task the uploader with going back to the RAW and reprocessing and re-uploading the photo?)

Where is the tool that remembers which photos have been opened or zoomed in and flags them as more interesting or worthy of backup?

Where is the tool that correlates the GPS location data of your iPhone photos and tentatively applies them to your corresponding camera photos?

Aperture used to collect photos from bursts into a single set and represent them with what it guessed was the best one. Where did this idea disappear to?

There’s a ton of low-hanging fruit here. Someone, please do something. I’m busy.

A Brief Foray into Random Name Generation

I got a bee in my bonnet about name generation this morning, so here’s a simple Javascript module for randomly generating names:

# NameGenerator


  const starNameGenerator = new NameGenerator(['Ceti Alpha', 'Scorpio', 'Draconis'...]);
  starNameGenerator.generate(); // -> random name

Works better with a decent sized (hundreds) of thematically similar examples to work from.
/* global module */

function pick(array) {
  return array[Math.floor(Math.random() * array.length)];

class NameGenerator {
// data is a map from character-pairs to observed successors,
// consider the examples "how", "now", "brown", "cow"
// the pair "ow" would have the following successors
// [undefined, undefined, "n", undefined] (undefined -> end of word)

  constructor(examples) {
    const data = {'': []};
    map(s => s.toLowerCase()).
    forEach(example => {
      let pair = '';

      for(let i = 0; i < example.length; i++) {
        pair = pair.substr(-1) + example[i];
        if (! data[pair]) data[pair] = [];
        data[pair].push(example[i + 1]);

    this.data = data;

  generate() {
    let s = pick(this.data['']);
    let next = pick(this.data[s]);
      s += next;
      next = pick(this.data[s.substr(-2)]);
    return s;

module.exports = NameGenerator;

I wrote a much more convoluted (but simple-minded) star name generator for my galaxy generator several years ago.  The approach I took was to take a collection of star names and break them up into syllables (starting, middle, and ending) and then given a range of syllable lengths, assemble a name out of random pieces.

Today it occurred to me that I’ve never explicitly implemented anything using Markov chains before, so how about I build something that way and see how it compares? If you follow the link, you’ll see examples of star names generated randomly the old way (names with “bad words” in them are rejected).

I took a list of proper star names from Wikipedia and cleaned it up (e.g. the Greek letter prefix of a star name simply indicates its brightness relative to other stars in the same constellation, while a Roman Numeral is simply an indicator of a star being part of a binary or trinary system). This gave me a bit over 600 star names with which to seed the generator, and the results are pretty nice. (Again, no need for bad world filtering.)

The major disadvantage of the new generator is that it can generate really long names pretty easily because of the way it terminates. If I implemented a more sophisticated generator that looked at things like overall length and length of current word in weighing the probabilities it would probably help here, but that might be overthinking it (it’s pretty easy just to reject overly long names).

In general, I think the new generator produces more pronounceable names than my earlier attempt, and some of the new names it generates seem like they should be real names, which is exactly what I’m hoping for.


The algorithm is very simple. The constructor looks at which characters follow a given pair of characters in the input data, so if you start with “how”, “now”, “brown”, “cow” you get the following data for the pair “ow”: [undefined, undefined, ‘n’, undefined]. So, using this data to generate names, 75% of names will end immediately after an ‘ow’ and 25% will have an ‘n’.

In this system, the first character of a name is following the empty string, while the second character of a name is following the first letter. It follows that using [“how”, “now”, “brown”, “cow”] all names will begin with ‘h’, ‘n’, ‘b’, or ‘c’. and most will end in ‘w’ and the rest will end in ‘n’. Not super interesting.

Let’s try slightly more interesting seed data: [‘Frodo’, ‘Peregrin’, ‘Meriadoc’, ‘Bilbo’, ‘Adalgrim’, ‘Bandobras’, ‘Celandine’, ‘Doderic’, ‘Erin’, ‘Griffo’]. This doesn’t seem like it will be enormously fruitful at first glance, but it immediately generates lots of new names that, to me, sound authentic: Adalgrin, Adobris, Grine, Bandine, Froderim.

And here’s a link to a jsfiddle to see it in action (with a bigger set of names from Middle Earth as the seed). One of the really nice things about it is that you don’t need to filter out bad words, because they pretty much don’t get created if they’re not in the source data.

It occurs to me that a lot of the random content generation stuff I’ve done in the past was, in effect, recreating Markov chains naively, and understanding what I’ve done in those terms is powerful and clarifying.

And with that, I leave you with a random Jabberwockish word generator. Don’t bewortle! And have a borpallith day.

Father’s Day — Picking Sides in Fallout 4

Random Prattle You Can Skip

As usual, I’m several years behind in my gaming, and — having pretty much completed Skyrim VR (all the major arcs, anyway) I decided to go back and try Fallout 4 (which had frustrated me around the first time I got my ass handed to me repeatedly by synths while doing nothing in particular).

father (head of the Institute)
Father — head of The Institute in Fallout 4 — and my excuse for writing this post today

I played Fallout and Fallout 2 back when they came out on the Mac in the 90s, and loved them. I played the hell out of Fallout 3, but got frustrated by New Vegas and never finished it. In general, I preferred the original games because you got to make significant moral choices (even if sometimes the content was juvenile).

Fallout 3 seemed to have turned into a more conventional action adventure, which I put down to the technical and cost limitations associated with AAA 3D games (building out huge alternate worlds to display widely varying outcomes is expensive, which is likely also why Mass Effect 3 was so lame.)

Anyway, Fallout 4 seems to me to be a return to the greatness of the originals. The moral issues are back and so, apparently are the choices. And, the game looks wonderful. (Both Skyrim VR and Fallout 4 have pretty terrible audio, frequently making me wonder if I have my headphones on backwards, which is impossible with the VR headset; there are a ton of physics, animation, and AI glitches — all of this just has to be accepted and ignored with Bethesda’s games.)

The Moral Question (SPOILERS!)

So, I’m pretty far along, and I’m confused.

In the Commonwealth there are several key factions, which I’ll summarize:

  • The Minutemen — these are the first folks you encounter, a rag-tag militia that’s almost been wiped out and simply tries to restore a cooperative community across the Commonwealth. Whether you like it or not you end up their “General” but (as someone put it eloquently on a forum) you still end up doing all their fetch and carry busywork.
  • The Brotherhood of Steel — this is the most obviously powerful faction. You encounter them, most likely, trying to find Diamond City (they’re very noisy!) Unlike the “lawful good” BoS from Fallout 3, these are more like the quasi-fascist BoS from Fallout 2. They make a good first impression (they’re defending themselves against feral ghouls), but if you steal a journal one of them has left lying around you quickly learn they have a dark side. In any event, when you join (as you pretty much have to) you quickly discover they are human supremacists and also pretty much anti-science.
  • The Railroad — a nod to the Underground Railroad in more ways than one, is a group of idealists trying to free synths from “slavery” and protect them from a surface population that hates and fears them. They’re run like a parody of John Le Carré.
  • The Institute — just as The Empire is introduced in Skyrim by trying to execute you for no particular reason, The Institute is introduced murdering your spouse (husband in my case) and kidnapping your son. Until you finally meet them face to face, all you know about them is that they killed your spouse, kidnapped your baby, they’re responsible for roving gangs of humanoid killer robots (that think out loud to somewhat comical effect), and that everyone blames them for everything, but disappearing people and replacing them in particular. Also, just as Skyrim’s Empire is the only place where you can bring up kids in relative safety and houses aren’t decorated with broken furniture and dead bodies, The Institute is the only place where you can sit down without wanting to sanitize yourself afterwards. (By way of role-play, I jumped on the chance to get a new toothbrush from a library vending machine…)

One of the interesting things I noticed in forum threads about the moral choice posed by needing to pick between these factions is (a) how sophisticated some of the arguments were (although there were plenty of “BoS are the only good guys” and “BoS are fascist losers” comments too), (b) how clever and effective Fallout 4’s writing is. In particular, in making their arguments, many players clearly take as fact that The Institute is guilty of all the things they’re accused of. Even though every missing persons case you investigate turns out to be entirely prosaic, everyone assumes that The Institute really is stealing people to use their organs, or something.

Here’s what you learn by actually visiting The Institute.

  • They create synths as servants.
  • They may have created the super mutants.
  • There doesn’t seem to be any overt cruelty in the way synths are treated, but then again they can mindwipe and reprogram synths at whim which is worse.
  • The gen-3 are almost certainly sentient, look completely human, and are clearly slaves, but The Institute’s citizens are in denial about this. (My favorite traveling companion is a synth detective, apparently a gen-3 prototype, who definitely seems to be sentient.)
  • You can see gen-3 synths being made from scratch out of goo via 3d printing. They’re not assembling stolen organs. (The sequence is very similar to Westworld’s credits, but this is in real-time game engine and came out before Westworld.)
  • It’s The Railroad that uploads people’s memories into synths to help them blend in (because otherwise the people on the surface would just kill them).
  • The Institute’s population includes scientists (“everyone here is a doctor, but I’m the only physician”) but also children (who seem totally normal and carefree).
  • The Institute have built a fabulous arcology underground.
  • Everyone in The Institute seems to be genuinely trying to make their own lives and life on the surface better. Yes, they’re ignorant and arrogant but so are the folks on the surface.

In a nutshell, the argument against The Institute is slavery / mindwipe. The argument for it is that it’s the only place you’d want to live or raise children.

Fallout 4 ultimately lets you choose between these four factions at the end. The Minutemen only really demand you kill The Institute whereas the other factions all want to wipe each other out (but don’t care about The Minutemen).

It seems to me that while The Institute is clearly wrong in its treatment of synths (i.e. slavery) it’s hardly worse than the Brotherhood’s position of exterminating all non-humans (i.e. genocide) especially when you consider that The Institute can actually make the world nice again and The Brotherhood can’t and won’t. It’s not at all clear to me that any of the factions are beyond redemption, but given that — uniquely in Fallout 4 — you’re someone who can remember the prewar world, it seems to me that there’s not much dividing the Brotherhood of Steel or the Minutemen from raiders and scavengers (indeed, there’s a diary in the room of the first synth you’re sent to recover detailing the collapse of a Minuteman splinter group into thugs and murderers), and the only way to return the world to a decent state is to keep The Institute around, and deal with its moral flaws.

Happy father’s day.