Lobsters: ShieldFont

Screen readers read the source code rather than the words displayed by the font, so the current <Shield> component uses aria-hidden to stop the substituted text from being read aloud. We have a built-in alternative, in beta and on by default, which allows users to spawn the real words on the page through an aria button only screen readers reach. The reader’s browser solves a compute-costly puzzle for the key, taking a few seconds. It works in our Dynamic Websites tier and requires JavaScript, which is part of the point: most scrapers never run any. Tested with VoiceOver and an automated screen reader.

Not a fan of making the experience even worse for screen readers.

This reminds me of how people tried to protect images from being copied and saved in the browser in the 2000's. Annoyances like prohibiting right click with JavaScript and overlaying the image with a transparent image (so that right clicking and 'Save as...' will save the transparent image) were widespread. It was completely futile.

This is an old technique students use to bypass plagarism software.

It's quite hostile to the end user as it will inevitably mess with your screen reader, clipboard, custom fonts etc.

Reportedly messes with screen readers: https://caneandable.social/@WeirdWriter/117031884001259326

Site now has a section where screen readers are excempt if they solve a "compute-heavy puzzle". So please ensure your screen reader has JavaScript!

This does nothing to prevent a scraper from overwhelming a site with traffic, though.

It does show that people are willing to jump through absurd hoops to mess with GenAI vendors though, which is interesting[1]

[1] or rather that the creator believes there's a market consisting of such people

Boy, I thought the tag was applied for the other reason. I can only assume the authors appreciate the ouroboric irony.

So... It is also applied for the other reason. This was partially vibecoded. Check the Agents file

Yeah, no kidding!

(I could have been less vague: reading the comments, I thought it got the tag for "mentions LLMs at least once", but lo, the site is almost caricaturistically performative and behold, the repo is dripping with vibes.)

Yeah, I got real whiplash from the purpose of the project and, well, everything else:

The design has all the hallmarks of Claude design sensibilities - the fonts, the bubbles, the colours, the cards-in-cards, etc
The prose is unnecessarily verbose and has a bunch of classic LLM tics ("What leaks is client code", "both fail silently", "no, and we won’t pretend otherwise", and so on)
The documentation for accessibility is the sort of gibberish that feels like poorly-reviewed, full-context glitches. Who or what is Sway in this context - the prompter?
The whole concept feels like AI in a way that I can't quite put my finger on, but there's something about the complexity, the tacked-on nature of accessibility (humans don't really tack on accessibility, they just don't do it in the first place?), or possibly just the way it's being sold as The Solution and not "here's a weird technique you could try".

That said, the link to Playtype does seem real, so presumably there's more than just a random developer behind this, throwing it together. It's disappointing they didn't put more effort into the details, though.

TL;DR: a special font gets generated that uses whole-word ligatures to show a different word (Rulers -> Authors, Ruler -> Husband, etc): https://i.imgur.com/AR1ElNz.png