A very astute framing by Ted Chiang—large language models as a form of lossy compression for text.
When we’re dealing with sequences of words, lossy compression looks smarter than lossless compression.
A lot of uses have been proposed for large language models. Thinking about them as blurry JPEGs offers a way to evaluate what they might or might not be well suited for.
A collection of stylesheets that don’t use class selectors. Think of them as alternatives to default user-agent stylesheets.
💯
I think it is beautiful if people have a purpose. But it should be valid to lead a purposeless life too. … Maybe it is okay to not pursue potential and just be okay with being.
It me:
Broadly, these are websites which are still web pages, not web applications; they’re pages of essentially static information, personal websites, blogs, and so on, but they are slightly dynamic. They might have a style selector at the top of each page, causing a cookie to be set, and the server to serve a different stylesheet on every subsequent page load.
This rings sadly true to me:
Suppose a company makes a webpage for looking up products by their model number. If this page were made in 2005, it would probably be a single PHP page. It doesn’t need a framework — it’s one SELECT query, that’s it. If this page were made in 2022, a conundrum will be faced: the company probably chose to use a statically generated website. The total number of products isn’t too large, so instead their developers stuff a gigantic JSON file of model numbers for every product made by the company on the website and add some client-side JavaScript to download and query it. This increases download sizes and makes things slower, but at least you didn’t have to spin up and maintain a new application server. This example is fictitious but I believe it to be representative.
Also, I never thought about “serverless” like this:
Recently we’ve seen the rise in popularity of AWS Lambda, a “functions as a service” provider. From my perspective this is literally a reinvention of CGI, except a) much more complicated for essentially the same functionality, b) with vendor lock-in, c) with a much more complex and bespoke deployment process which requires the use of special tools.
A potted history of communication networks from the pony express and the telegraph to ethernet and wi-fi.
I like this high-level view of the state of CSS today. There are two main takeaways:
This is exactly the direction we should be going in! More and more power from the native web technologies (while still remaining learnable), with less and less reliance on tooling. For CSS, the tools have been like polyfills that we can now start to remove.
Alas, while the same should be true of JavaScript (there’s so much you can do in native JavaScript now), people seem to have tied their entire identities to the tooling they use.
They could learn a thing or two from the trajectory of CSS: treat your frameworks as cattle, not pets.
Even if you can somehow justify using tracking technologies (which don’t work reliably) to make general, statistical decisions (“fewer people open our emails when the subject contains the word ‘overdraft’!”), you can’t make individual decisions based on them. That’s just wrong.
Elise Hein documents what it was like to build a website (or web app, if you prefer) the stackless way:
- use custom elements (for modular HTML without frameworks)
- use the in-browser package manager (for JavaScript packages without build tools)
- match pages with files (to avoid routing and simplify architecture)
- stick to standards (to avoid obsolescence and framework fatigue)
Her conclusions are similar to my own: ES6 modules mean you can kiss your bundler goodbye; web components are a mixed bag—it’s frustrating that Apple are refusing to allow native elements to be extended. Interestingly, Elise feels that a CSS preprocessor is still needed for her because she wants to be able to nest selectors …but even that’s on its way now!
Perhaps we might get to the stage where it isn’t an automatic default to assume you’ll need bundling, concatenation, transpiling, preprocessing, and all those other tasks that we’ve become dependent on build tools for.
I have a special disdain for beginner JavaScript tutorials that have you run
create-react-appas the first step, and this exercise has only strengthened my conviction that every beginner programmer should get to grips with HTML, CSS and vanilla JS before delving into frameworks. Features native to the web are what all frameworks share, and knowing the platform makes for a stronger foundation in the face of change.
Many, if not all, of our world’s most wicked problems are rooted in the excessive hiding of complexity behind illusions of simplicity—the relentless shielding of messy details in favor of easy-to-use interfaces.
But there’s always a tradeoff between complexity, truth, and control. The more details are hidden, the harder it is to understand how the system actually works. (And the harder it is to control). The map becomes less and less representative of the territory. We often trade completeness and control for simplicity. We’d rather have a map that’s easy to navigate than a map that shows us every single detail about the territory. We’d rather have a simple user interface than an infinitely flexible one that exposes a bunch of switches and settings. We don’t want to have to think too hard. We just want to get where we’re going.
Seamful and seamless design are reframed here as ethical and deceptive design:
Ethical design is like a glove. It obscures the underlying structure (i.e. your hand) but preserves some truth about its shape and how it works. Deceptive design is like a mitten. It obscures the underlying structure and also hides a lot about its shape and how it works.
I really enjoyed this trip down memory lane with Chris:
From the Web’s inception, an ancient to contemporary history of the Web.
This is a superb twenty minute presentation by Trys! It’s got everything: a great narrative, technical know-how, and a slick presentation style.
Conference organisers: you should get Trys to speak at your event!
“Serverless”, is a buzzword. We can’t seem to agree on what it actaully means, so it ends up meaning nothing at all. Much like “cloud” or “dynamic” or “synergy”. You just wait for the right time in a meeting to drop it, walk to the board and draw a Venn Diagram, and then just sit back and wait for your well-deserved promotion.
That’s very true, and I do not like the term “serverless” for the rather obvious reason that it’s all about servers (someone else’s servers, that is). But these three principles are handy for figuring out if you’re building with in a serverlessy kind of way:
- You have no knowledge of the underlying system where your code runs.
- Scaling is an intrinsic attribute of the technology; so much so that it just happens automatically.
- You only pay for what you use.
Abstraction; scale; consumption.
Lessons for web development from a home renovation project:
- Greenfield Projects Are Everyone’s Favorite
- The Last Person’s Work Is Always Bewildering
- It’s All About the Trade-Offs
- It ALWAYS Takes Longer Than You Think
- Communication, Communication, Communication!
And there’s this:
You know those old homes people love because they’re unique, have lasted for decades, and have all that character? In contrast, you have these modern subdivision homes that, while shiny and new, are often bland and identical (and sometimes shoddily built).
node_modulesis like the suburbia/subdivision of modern web development: it seems nice and fancy today, and most everyone is doing it, but in 30 years everyone will hate the idea. They’ll all need to be renovated or torn down. Meanwhile, the classical stuff that’s still standing from 100 years ago lives on but nobody seems to be building houses that way anymore for some reason. Similarly, the first website ever is still viewable in all modern web browsers. But many websites built last year on last year’s bleeding edge tech already won’t work in a browser.
The Web is smothering in useless images. These clichéd, stock images communicate absolutely nothing of value, interest or use. They are one of the worst forms of digital pollution because they take up space on the page, forcing more useful content out of sight. They also slow down the site’s ability to download quickly.
Chris broke both his arms just to avoid speaking at the JAMstack conference in London. Seems a bit extreme to me.
Anyway, to make up for not being there, he made a website of his talk. It’s good stuff, tackling the split.
It’s cool to see the tech around our job evolve to the point that we can reach our arms around the whole thing. It’s worthy of some concern when we feel like complication of web technology feels like it’s raising the barrier to entry
Twenty hard-won lessons from Dan from ten years of Dribbble.
We sent 50 shirts along with a card to friends and colleagues announcing Dribbble’s beta back in 2008. This first batch of members played a pivotal role in the foundation of the community and how it would develop. The shirt helped guilt them into actually checking out the site.
I think I still have my T-shirt somewhere!
Tom makes an endpoint for generating QR codes so you don’t have to rely on the Google Charts API.
He also provides a good definition of “serverless”:
Now, serverless is a very silly buzzword dreamed up by someone from the consultant class who love coming up with terrible names, so I promise I won’t use it any further. Your code obviously run on a server. It just means it runs on a server someone else manages.
Amazon call it a ‘Lambda Function’. Google call it a ‘Cloud Function’. Microsoft Azure call it simply a ‘Function’. But none of those are very descriptive, because, well, anyone who writes any kind of programming language generally writes functions pretty much all the time in much the same way as anyone who writes English writes paragraphs, and we don’t call our blogging software “Cloud Paragraphs”. (Someone will now, I’m guessing.)
Here’s a clever tiny lesson from Dave and Brad: you can use prefers-reduced-motion in the media attribute of the source element inside picture.
At Clearleft, we’re always saying “Everything is a tiny lesson!”, so I love, love, love this bit of Manuel’s website where notes down short code snippets of little things he learns.
Of all the buzzwords in tech, perhaps none has been deployed with as much philosophical conviction as “frictionless.” Over the past decade or so, eliminating “friction” — the name given to any quality that makes a product more difficult or time-consuming to use — has become an obsession of the tech industry, accepted as gospel by many of the world’s largest companies.