Jack and I prove that humans are not web scale

Of course, in a Twitter thread

(Vicki’s Note: I’m going to do one more post after this one for 2019 ( fingers crossed), and then Normcore is on vacation for the rest of the year. I’ll be back next year with more content, and, hopefully Normcore laptop stickers!)

There is a very old (by modern web standards) video floating around the internet, called “MongoDB is Web Scale”. It was recorded by using Xtranormal, a software that turned written scripts into dialogues read by cute, puffy cartoon characters. These kinds of videos were in vogue around 2011, and the crux of the humor rests partially in the robotic presentation, and partially in watching pastel bears and cats ranting about sharding.

In short it’s about one animal deciding to try the then-controversial MongoDB for its website’s backend because it had read somewhere that the NoSQL database was web scale, while the other animal disgustedly tries to talk it back to reason. Eventually, the reasonable animal fails, gives up, and goes to live on a farm to clean up after pigs because that life is preferable to dealing with arguments about software architecture.

I didn’t link to it because the video itself is pretty vulgar and this is a PG-13 rated newsletter at best (except for when I write about Paul Graham, in which case it becomes rated PG), but you can find it very easily online. The trademark phrase throughout is “MongoDB is Web Scale,” and I’ve been thinking about this phrase as we reach the end of the year and I rifle back in my mind through all the Normcores of months past.

What does it mean for a technology to be web scale? It means a website or app can take on an ever-increasing amount of traffic by distributing the load across multiple machines. Examples of web scale companies are all the usual suspects: Netflix, Google, Amazon, and Twitter.

It’s not easy to make systems web scalable. There are all sorts of security concerns, transaction checking, and alerting that needs to happen as the amount of computers you involve in computations grows and grows and grows.

As of 2019, having overcome a ton of security and architecture issues (lots of fun stories in this link), Mongo is pretty much web scale.

However, humans are not, and will never be. The modern internet we’ve invented for ourselves works great for the millions of machines we keep adding, but is a hellscape for people to navigate because it goes against every inclination of human nature.

One of the biggest problems with today’s internet, and which This Newsletter does a valiant job trying to fight every week, is that we are dis-incentivized to synthesize information at every turn.


Synthesis is one of the hardest mental tasks we as humans can possibly do, because it takes a lot of energy and concentration to gather all the information we know, scramble it together, and come up with an aggregate opinion. It’s the human equivalent of map-reducing data, and there’s a reason people build whole energy-hogging web-scale cloud platforms to aggregate and analyze all of the raw log data (Normcore link) that we get. It’s also the most important for humans to make decisions on and to understand our world.

Donald Norman, in his seminal 1993 book, “Things That Make Us Smart: Defending Human Attributes in the Age of the Machine” (which you should read!), says the same thing. He says there are two types of cognition: experiential and reflective. Experiential cognition means doing something quickly, based on experience. He writes that this type of cognition results in, “decisions that require considerable insight and information can thereby be made rapidly and without apparent thought.” He gives an example of an airline pilot reacting to a change in the weather. Although the pilot has years and years of training, he makes his decisions instantaneously, reacting to the environment based on instinct.

Reflective thought though,

is very different from experiential thought, even when both are applied by the same people in similar situations…Reflective thought requires the ability to store temporary results, to make inferences from stored knowledge, and to follow chains of reasoning backward and forward, sometimes backtracking when a promising line of thought proves to be unfruitful. This process takes time. Deep substantive reflection therefore requires periods of quiet, minimal distraction.

And if there is only one thing I’d love for Normcore readers to take away, after over half a year of the newsletter (other than the fact that you probably shouldn’t use Kafka, invest in crypto, or visit WeWork), is that it is this kind of reasoning, thoughtful reflection, and delayed reaction that is most important for everyone who lives online. But it’s even more critical to people who work with information and shape the decisions we make today.

However, today’s internet is built for web scale, but it is not built for this kind of thought process. In fact, everything online is actively meant to discourage pausing and thinking. And, because hundreds of thousands of people have spent years thinking about how to get us to click, it is almost impossible for us to fight against It. For example, in spite of my best efforts and even writing a newsletter about not digging beneath the online hype, I recently fell into the experiential cognition trap myself.

Jack and I

It happened like this. A couple weeks ago, after a protracted process of soul-searching and silent retreats in Myanmar (Normcore link), Jack Dorsey announced that, partially in response to all of these problems with today’s internet, he wants to create a decentralized Twitter.

Why? He wrote a long thread about the difficulty of moderating a large platform, and about the benefits and challenges of doing so.

Almost as soon as the thread appeared in my timeline, people started reacting to it, either with one-liners about how important it was, with jokes, or with criticisms.

The nature of Twitter is that you have to think fast and weigh in on the zeitgeist immediately, or your voice gets lost forever in the shuffle, so I skimmed the thread and decided to weigh in, without going back and doing the synthesis.

I could have researched Twitter’s past attempts to become open, and tweeted that Twitter had already had a chance to participate in a decentralized version of itself, through its third-party APIs, which it closed to developers after many years of poor relationships.

I could have looked into Jack’s incentives for pushing this platform, gone back and read about his recent involvement and influences in blockchain, particularly contrasting it with his other company, Square’s involvement in crypto (which would have been easier for me to understand now that I’ve written a whole thing about it.)

I could have read up on decentralized internet systems in general and weighed in on whether they were effective and profitable yet. (Answer: not really)

But all of that would have taken time, and people were already slinging hot takes, and the tweet was going down farther and farther on my timeline, and I needed to say something immediately. I was desperately fighting against the human impulse to be consulted, to weigh in, amplified by the feed, and I lost.

So, instead of something nuanced and smart, I wrote a tongue-in-cheek tweet about a three-billion dollar company hiring five people to fix the internet:

And, naturally, a couple hours later as happens quite frequently on today’s internet, the person who the sub-tweet was about saw the tweet. Jack replied,

“Rich, dynamic, continually present environments can interfere with reflection. These environments lead one towards the experiential mode, driving the cognition by the perceptions of event-driven processing, thereby not leaving sufficient mental resources for the concentration required for reflection,” writes Norman, and nowhere is this more true than here.

First, Jack’s initial post about the new decentralization team, Bluesky, was split up into a thread. Twitter threads are super hard to read and extremely ephemeral - people delete messages, threads are hard to search for and through, and they get updated in real-time, where the latest tweet gets pushed to the top of your timeline, regardless of the rest.

I get that it behooves Jack to use his platform as it was designed by the people he told to design it, as a signal of confidence. But wow was it hard to read that thread and not skip parts. Coincidentally, I don’t see the announcement anywhere on Twitter’s official website.

Second, as a result of the fast-moving, knee-jerk nature of Twitter, both Jack and I felt compelled to respond in a sub-optimal way. Just as I didn’t give a response that had any research in it, so did he. What did “ha” mean? Did Jack think it was truly funny? Was he laughing along? Was he hurt by my tweet? Did it mean he was going to cancel me? (Luckily no, and I’ve been able to continue shitposting about Python…so far.)

Both of these things, as well as the larger issue, mean that Jack’s establishment of the Bluesky team he talked about in the tweet is of utmost importance. He talks about then need to shift the focus of Twitter from away moderation, open recommendation algorithms, allowing for healthy conversation, and all of great stuff I’ve been championing here on a regular basis: small internet, more humans in the loop, and insight into black box machine learning algos. We need to step back from web scale, at least a little.

Pandora’s Internet

Because, in addition to creating problems on Twitter, web scale has opened up a whole other Pandora’s box. Not only does it lead to a very shallow dialogue on important issues, it has, basically, fried our brains and resulted in some of the worst outcomes for humanity.

Deepfakes. Targeting children on Instagram. Cancel culture. Bad YouTube recommendations for kids. Constant surveillance culture. Those same video cameras getting hacked. The fall of journalism and the rise of Buzzfeed. Mass Whatsapp hysteria. The fall of creativity online. The spread of anti-vaccination information on social media platforms. An increase in anxiety and depresion due to the use of social media. Facebook moderators dying on the job. Apps circumventing your privacy settings. No file ownership. Scammy Amazon reviews. Yelp extortion. Gaming Google results. Data rot. Bad self esteem on Instagram. Weird Brand Twitter. Social Credit Monitoring. And, of course, TayBot.

Why is it like this? We took this beautiful thing with so much promise and trashed it. Why are WE like this?

Because we as people are terrible at grasping web scale in our minds. People cannot scale up like machines. We can’t possibly think through all of the consequences of creating behemoth sites and systems that impact billions of people. No one has so far, and no one will in the future.

It’s just easier if there’s less web to go around - or at least - have the web giants be able to reach and be responsible for less people simultaneously.

In addition to Twitter’s efforts, there are some examples of technologies that emphasize this human-scale approach. For example, Postlight (founded by Paul Ford, who wrote “Why Wasn’t I Consulted”), recently released a fun side project called Yap, ephemeral chat for six people. (Remembering is for machines, the right to be forgotten is important for humans.)

Substack is also what I consider to be a human-scale company in some ways: you can have a newsletter reach hundreds of thousands of people, but on a one-on-one basis, and in order to talk, they have to reply to you directly. DuckDuckGo doesn’t track your search history. Telegram (Normcore link) is not affiliated with any major chat service and is great for small group chats.

Mastodon is already decentralized Twitter. Keybase, now that it’s starting to get their act together, is great for secure chat.

The more examples like this we see and the more examples we see actually being financially viable, the more we can move away from the hideous, gargantuan of web scale, and back towards human scale.

Even though our Twitter exchange was terrifying, I am optimistic that the CEO of a leading internet platform thinking about decentralization means something big is coming. Or rather, something small. The last ten years have been about big data and web scale (Normcore link).

The next ten will, if everything goes right (and, being Eastern European, I can’t say with confidence that it will - but it’s all in our hands), be about medium-sized data and people.

Art: Crowded Boat II, Robert Goodnough, 1963

What I’m reading lately:

  1. Journey into how e-waste gets recycled

  2. The GitHub Typo Corpus

  3. This podcast about Dolly Parton and what her career says about America

  4. This list of predictions for the next year, including mine

  5. The creator of Nginx is in trouble

  6. Leggings at work

    About the Newsletter

    This newsletter is about issues in tech that I’m not seeing covered in the media or blogs and want to read about. It goes out once a week to free subscribers, and once more to paid subscribers. If you like it, forward it to friends!

    Select previous free Normcore editions:
    Keybase and the chaos of crypto · What’s up with Russia’s Internet· I spent $1 billion and all I got was this Rubik’s cube· Die Gedanken sind frei · Neural nets are just people· Le tweet, c’est moi· The curse of being big on the internet· How do you like THAT, Elon Musk?·Do we need tech management books? ·Two Python Paths
    Select previous paid Normcore editions:
    Sidewalks for the internet· Imgur is bad now · Eric Schmidt and the great revolving door· No photos please · Deep thoughts of Cal Newport

    About the Author:
    I’m a data scientist in Philadelphia. Most of my free time is spent wrangling a preschooler and a baby, reading, and writing bad tweets. I also have longer opinions on things. Find out more here or follow me on Twitter.