Skip to main content
Baldur Bjarnason

Book production, AI, Single-Page-Apps, and other links and notes

Baldur Bjarnason

Some thoughts on how to make a book, three months after I made one #

Over on my personal blog I wrote up a few notes on the approaches I took when I made my book. Contains advice on editing, covers, that sort of thing.

The open-source HTML+CSS to PDF tools that are available are now definitely good enough for these sorts of projects.

I might dig a bit more into the print stylesheet I used at some point, if people are interested.

You can find the crufty, unedited, and messy current version over here.


“<3 Deno” #

I’ve been using Deno quite a bit on a project and it seems to generally be full of sensible approaches and common-sense decisions. Also, I’ve noticed on social media that a few of the people involved in npm seem to think that Deno is making huge mistakes with its module system and dependency management.

This is probably a point in Deno’s favour, to be honest.


 “Machine learning offers a fantastically powerful toolkit for building complex systems quickly. This paper argues that it is dangerous to think of these quick wins as coming for free.”  #

Integrating machine learning features with normal software systems looks tricky. This is a 2014 paper that goes over some of the issues.


 “Inside the Heart of ChatGPT’s Darkness”  #

“OpenAI’s current guardrails are only skin deep; some serious darkness still lies within.”


 “Automated testing won’t solve web accessibility - Eric Eggert”  #

While I think automated testing tools are not good at finding accessibility issues and helping to remediate them, the tests can help with keeping a website or app accessible. Significant issues will crop up earlier when you have regular linting and automated testing for accessibility. This becomes even better if you can include best practice rules. Comparing to a known best practice will always be more useful than having a general purpose tool.

Automated accessibility testing is useful but not sufficient.


"The Limits of ‘Computational Photography’ - PetaPixel" #

So, because tech is a pop culture, it will apply techniques and approaches because they are fashionable even when the outcomes are worse.


Extraordinary claims require extraordinary evidence, and AI/ML research isn’t providing it #

​Whenever you see an AI/ML paper making extraordinary claims remember that the field has a reproducibility crisis due to poor modelling, evaluation, over-optimism, and data leakage

“Leakage and the Reproducibility Crisis in ML-based Science”

Moreover, the field has suffered from this issue for a while

This crisis may be hard to address because the ML models are black boxes. For many, it’s essentially impossible for researchers to be sure that there isn’t data leakage.

​And given the size of the training data used in upcoming versions of these models, data leakage seems inevitable. Unless you get really creative, it’ll be very hard to come up with tests and experiments that aren’t variations of data already in the model

Remember, extraordinary claims require extraordinary evidence.


AI Everything, Everywhere! #


For those committed to ignoring history: “the brain works just like this new tech” is a recurring theme in tech discourse and it’s always wrong

The brain does not work like a clockwork mechanism, transistors, computers, software, or LLMs.

The brain, so far, is a unique thing.

And, no, making the analogy more specific does not make it more accurate. Human creativity does not work like a Large-Language-Model. Creativity is a process of reasoning. Some of it is ad hoc. Some of it is emotionally motivated. But it’s a form of reasoning through and through.


My CSS wishlist is simply to have more time to be able to do things properly.

Also, my JS, HTML, and blogging wishlist.


Accidentally reset the zoom in Visual Studio Code and, wow, the defaults are so clearly set by somebody in their 20s.

Visual Studio Code is the only app on my desktop where I need to boost up the font size from the default.


"Why I’m not the biggest fan of Single Page Applications - Manuel Matuzović" #

There have been a few of these lately, for good reason. Single-Page-Apps are a generally ill-considered practice that’s finally falling out of fashion because cost-cutting has become the norm and developer influencers have a strong preference for minting a New Popular Thing over just plain losing their jobs.


Single-Page-Apps as a Symptom of Pop Culture #

“Single-Page-Apps are bad” discourse has resurfaced because, y’know, Single-Page-Apps are kinda bad, all else being equal.

Drama ensues because if SPAs are bad for most websites and people knew that, then somebody must have misled somebody else.

My take on that is that tech cos and software development are a pop culture that does things because they are fashionable. SPAs were en vogue, therefore they were the default. There was no malice because tech is driven by fashion, not outcomes or cost.

I’ve written about this pop culture a couple of times:

The idea that tech is a pop culture comes from Alan Kay:

Although he is less fond of pop culture in general than I am. He also doesn’t talk about the role of counterculture in a sphere dominated by pop culture. Criticisms and reactions to a mainstream pop culture generally fail to change the mainstream but instead lead to the formation of a counterculture that defines itself in opposition to the mainstream. They also serve a role in the mainstream culture: they define its edges by providing examples of outsiders and outside opinions. Punk versus Pop, if you will.

If a practice in a pop culture is fashionable, then there are no significant downsides (or they get reframed as benefits) and everybody who says otherwise is an unfashionable hater who belongs to the counterculture. Pop cultures are self-reinforcing and the only way to win is not to participate and focus on keeping your own corner of the industry as rational as possible.

Software development culture can’t be fixed with discourse.

(Though, mind you, I have been accused of pessimism)

One consequence of this is that the discourse under these circumstances often takes on the characteristics of a morality play where the sides are mostly just sharpening their own opinions instead of actually debating.

I’ve also written about this in the past:

Calling web dev a pop culture isn’t a defence. It is a description, through analogy, of the mechanisms and the dynamic of the web dev industry that disconnects practice from the direct consequences of bad outcomes. It is a model that helps explain why bad practices thrive. It’s a framing intended to help people mitigate the harm, not to excuse it.


"Automating podcast transcripts on my Mac with OpenAI Whisper – Six Colors" #

Making audio transcripts easier is one of the constructive uses of this tech.


“Oops! How Google bombed, while doing pretty much exactly the same thing as Microsoft did, with similar results” #

“AI has a fairly spotty track record for taking demos into reliable products”


That 90s Show feels like one of those show-within-a-show bad sitcoms that’s included in a middling sitcom about how dysfunctional TV production is.


I’m not the only one to doubt the legality of ChatGPT and Large-Language-Models under the GDPR.


“Nintendo to increase wages 10% despite lowered forecast” #

A few days ago, I posted a link to a 13-year-old story about Nintendo refusing to do layoffs despite a recession. Looks like its management culture hasn’t changed.


“Should Schools Tackle Andrew Tate Head On? - Novara Media” #

​Clara and John are among a growing number of teachers who are concerned about the influence of the former kickboxer, entrepreneur and social media personality Andrew Tate. The self-styled “king of toxic masculinity” is currently imprisoned in Romania on suspicion of rape and sex trafficking, allegations he denies. Although he’s now banned from all major social media platforms because of his sexist content, his viral videos continue to circulate. On TikTok, videos of him have over 11 billion views.

This Week’s Obsessive Listening #