this is not something I came up with, Simon wrote it and I liked the differentiation between "vibe coding" where there is less effort
for this case project I think I would actually go back and say it's vibe coded, but I didn't want to just call it vibe coding because I did spend time going back and forth and directing the agent
Interesting distinction. I've previously heard vibe coding described as "vibe prompting, but you actually do some work." That aside though, I just call what you're describing as coding with AI.
coding with AI is coding just as much as coding with VSCode is coding. you decide which parts you get help from a given tool and which you don’t. end of the day, it is all coding and “coding with AI” sounds as silly as “coding with keyboard / microphone”
The first part is exactly my point, but the latter is nonsense in my book. You cannot ask VSCode (pre-AI) to write a program for you. It's akin to doing math with AI vs. an Nspire CAS. There's no reason to think you need to respond to those who shame vibe coding with claims that we shouldn't differentiate our tools, but we also shouldn't just say it's all the same. We wouldn't claim that about farming with a laser-powered weed killer compared to farming with a horse-drawn plow.
For software, but that's a well trodden path at this point. I've seen a few projects that are actually "vibe engineering" outside of software on the 3d modeling side so the terms are confusing.
thanks for the info, I'll see if I can get a agent to fix it
it's a static webpage, the source is available with right-click view source, I added a BSD2 licence header to it to make clear it's fine to take and do mostly whatever with
Looks useful but doesn't work quite as expected for me.
In Vivaldi location tracking doesn't work.
Version
7.7.3851.66 (Official Build) (64-bit)
Chromium Version
142.0.7444.245
Extended Stable channel (may also include additional security patches)
Channel
Official Build
Platform / OS
Linux - linuxmint 21.3
And in Firefox 146.0.1 on the same machine the URL doesn't get updated.
And if you are open to bug reports.. if I move around the drawings move smoothly with the map, but if I zoom in/out the drawings move only after the map zooming animation ends, rather than smoothly
But not well tested. Try to create a map and copy the url to another map. Now change the first map with more anotations or move the map center and copy the generated url and paste it into the other map on the other browser. That does not work (at least for me on different browsers).
I think I know what you mean, thanks for the report, if you modify the # part on a webpage it's not the same as reloading it, and I doubt I watch for that part changing
yeah, isn't it impressive how fast modern computers can be if you make a bit of effort, in this case I think I told it to just use plain javascript and make sure it's fast :-)
it's a static webpage, the source is available with right-click view source, I added a BSD2 licence header to it to make clear it's fine to take and do mostly whatever with
Per the spec [0], a URL can hold at least 8,000 characters.
> It is RECOMMENDED that all senders and recipients support, at a minimum, URIs with lengths of 8000 octets in protocol elements. Note that this implies some structures and on-wire representations (for example, the request line in HTTP/1.1) will necessarily be larger in some cases.
Mainstream browsers support at least 64,000 characters [1], and Chrome supports up to 2MB [2].
One thing I've learned from checking up on assumptions I've had about history is that it's easy to underestimate people in past times. They were probably better at communicating this stuff than you think.
I can open the page with the book text on mobile Safari, but iOS seems to cut off content when trying to copy/share the page URL. I can't get it to survive a round-trip to Notes. Might be a good thing to note for mobile users that if they write too much attempting to save their link will corrupt it.
This unfortunately immediately crashed my android firefox nightly browser. Amusingly it loaded the page, but one click on the address bar sent me straight to the home screen
I guess the surveillance industry has enough incentives to make this ever larger, so they can fit more utm-trackers, campaign-ids, referal trackers and whatnot in URLs.
It's truly insane how large typical share-URLS for content on instagram, youtube or any other large platforms are. URLs that could've been example.com/t/some-large-enough-id?time=13337 are stuffed with hundreds of characters, just to gather more data on people using these links.
> Per the spec [0], a URL can hold at least 8,000 characters.
> It is RECOMMENDED that all senders and recipients support, at a minimum, URIs with lengths of 8000 octets in protocol elements.
It is always worth remembering that, unless you have already ensured that the content has been rendered into a URI-safe subset of ASCII, a character and an octet are not the same thing.
Was just working on something similar this morning. As an fyi you can avoid the string replacing in the base64 string by using `.toBase64({ alphabet: "base64url" })` and `fromBase64({ alphabet: "base64url"})`.
I am thinking from a piracy perspective. If I share a link that contains a book, what can be done from DCMA or legal regulators? They can't ask the server (textarea.my) to remove the link because it doesn't exist.
They can't track every website with the link and ask to be removed, either.
Could they ask textarea.my to not parse the link and thus, not display the content? Could textarea.my refuse?
From a technical perspective, you're absolutely correct.
From a regulatory perspective, it seems unlikely that most courts would appreciate the difference. In their mind - you run a website, and that website contains copyrighted content. Take it down.
You'd probably have to just blacklist the link in question to avoid a legal headache.
In this case I'd say the link is the content. So it would be the place where you share the link, rather than the "rendering page", which should be more worried
Someone in another comment posted Crime and Punishment by Fyodor Dostoevsky, so a book fits in the URL. It is just that the URL is ~500000 characters :-P (the book itself is ~1.2M characters)
I recently build a small framework to create JavaScript apps that use this kind of URL sharing and therefore don’t require a backend: https://github.com/grothkopp/lost.js
By the look on the issues there, it seems the rest of the post is not that true either
Edit. Call me a hater, but... I know the guy! That's the guy from Google whose code never works in the most hilarious ways! See issues on the rest of his pinned repos.
I really like this from a privacy point of view. So much so that I'm thinking about adding a purely URL-storage solution as an option in my https://kraa.io editor.
From purely privacy point of view it’s not. But if you also want markdown features, custom typography and easy sharing, this starts to make more sense.
I made something similar once, specifically targetted for guitar tablature https://tabviewer.app/
To make links shorter for sharing with others, I use a shortlink service. Pasting URLs of thousands of characters long can be problematic
Think you've inadvertently found a way to provide extra tests for mobile devices.
The Crime and Punishment one consistently crashes Brave mobile for me. I assume it's the length of the URL - and seen another commentator say the same for chrome mobile (sure they both use the same codebase so likely an upstream issue).
Can anyone think of a way to store the textarea value in the URL? I tried using JS to set a # but it's nonsensical in this context.
Edit: here's the best I could do:
data:text/html,<title>Notepad</title><textarea id=t autofocus spellcheck=0 style=position:fixed;inset:0;padding:2em;border:0;font:monospace></textarea><a id=s style=position:fixed;top:10px;right:10px>Right-click Open to save...</a><script>[,P,S]=location.href.slice(15).match(/(.<textarea[^>]>)[^]?(<\/textarea>.)/),t.oninput=U=_=>s.href='data:text/html,'+P+encodeURIComponent(t.value.replace(/&/g,'&').replace(/<\/textarea/g,'</textarea'))+S,U()</script>
My focus was on finding a good text→URL-slug compression strategy. I used ChatGPT-5.2-Pro mainly to explore and compare different compression approaches and trade-offs.
I love this. Great little html page to refresh on Javascript.
For fun I put it in chatgpt and asked if there are bugs.
It warned about fromBase64() and toBase64() not existing in main browsers. It is supported but is indeed a new "baseline 2025"feature. It suggested more compatible code using two small functions to convert characters manually.
"deflate-raw is not consistently supported." It suggested using 'deflate' instead.
A few weeks ago I vibe coded a guitar tab editor just because I wanted to share a quick tab in a chat group with my band. When the first prototype already worked great, I just couldn’t stop to add features so that it now even has mouseover chord diagrams and copy and paste.
The sharing works just like here, by encoding the tab itself in the url.
caveat emptor re long hashtag techniques on (ipad) safari ...
you may think safari has no effective url limit (i.e. very high) but if you ever treat a url within the url bar as editable you are at risk to be silently truncated to 4096 bytes (eg select a character in the url bar and replace it)
also re-testing potential ~buffer limits in various ways on ipadOS 26.2 safari just now slowed my safari ui down to a crawl
eg after saving example.com with ~20k #hashtag to reading list -- each keystroke in this reply was taking several seconds, so I had to force quit safari and retype to post this warning
Students are lazy, in a good way, so they are more likely to run things on their own and play with interactive bits if the whole lecture is just one link.
Not really… using js to change the CSS on the go is not a good practice. Why does it matter? Because of the “dark mode” browser extensions. They often use the presence of @media query (or other standard CSS means of setting dark mode colors), and if it’s the JS that changes the colors we often get partial Dark Mode, which does not work at all.
I like these kinds of projects, but adding a file export/import is inevitable. It's less about the limits of a URL and more about practicality.
I also have no way to confirm that URLs aren't logged server side, so I'd never trust the claim about "no tracking". That's why these projects also end up self-hosted.
Typos and URL mangles are common though, and I'd still have no way to confirm if it got logged in that case. It's out of scope for anything in the github source, and instead depends on the server hosting the page. I know this isn't meant to be super secure, but it's still worth a mention.
Typos aren't making the hash part turn into something else. Like your parent comment explained to you, the hash part is not sent to the server. If you go out of your way to mangle the URL then of course a mangled URL without hash will likely get logged to the server. But I'm not sure how one would manage to go so much out of the way that they mangle the URL in a way that removes the hash.
You don't have a choice pasting links into some apps. They may strip out query and hash components, percent encode, force URL shortener services, etc.
Percent encoding is particularly bad since it may also bloat the length causing truncation and the decompress to fail. There's endless footguns with URLs.
Very nice exploration of URL-as-state. The approach is elegant, but the mobile crashes highlight how hostile real-world URL handling still is once links leave the browser.
Thanks for sharing! I tried a similar content-in-url approach for a family grocery list app but I couldn't get the url that short. (It worked but it was a bit cumbersome sharing over Whatsapp.) Will see what I can learn from this!
I created a similar app just 2 days ago targeting Whatsapp (https://linqshare.com) . Context: In my locality, EA, we normally have Whatsapp groups raising funds for whatever reason; for every content edit, the admin has to copy-edit-paste updated content(which contains name and amount) to the group. This small app intends to provide a table that's easy to convey this info. App stores content in the url but a preview image (needed for Whatsapp share) is stored at R2. Let me know if you want the source code running at Cloudflare.
In Firefox, https://textarea.my shows up as as a completely static non-actionable white page. Just white, with default cursor. No errors on the console.
I wonder if this can be paired with a local URL shortener? Chaining this with a local URL shortener can mean access to any doc with a single letter (or very letters).
I think a couple of days ago I stumbled upon your editor in corp Google intranets when I was looking for internal tool to pretty print some json, small world :)
love it, funny enough, I had similar idea pop into my head some weeks ago, just to be able to store quick notes and favorite them in my browser for later
I needed a way to share a link to a map, with drawings and the ability for the receiver to see their own location on the map.
Annotated screenshots solves the first but not the second.
Vibe engineered this, with many of the same ideas as OP.
Took an evening. Just in time apps for one specific use case is a thing.
And because it's so cheap to make and can be hosted cheaply with no backend, it can be given away for free.
https://nyman.re/mapdraw/#l=60.172108%2C24.941458&z=16&d=LU8...
While I'm all for vibe coding as appropriate, there's a lot of humor to be found it calling it engineering. :D
for this case project I think I would actually go back and say it's vibe coded, but I didn't want to just call it vibe coding because I did spend time going back and forth and directing the agent
https://simonwillison.net/2025/Oct/7/vibe-engineering/
Is the code open source online somewhere?
it's a static webpage, the source is available with right-click view source, I added a BSD2 licence header to it to make clear it's fine to take and do mostly whatever with
Here is the fix:
.leaflet-top, .leaflet-left{ z-index: 100000; /* some high number */ }
although I run 140.6.0esr so maybe newer ones need a even higher one?
the code is on GH now https://github.com/gnyman/mapdraw , codeberg is on my todo
In Vivaldi location tracking doesn't work. Version 7.7.3851.66 (Official Build) (64-bit) Chromium Version 142.0.7444.245 Extended Stable channel (may also include additional security patches) Channel Official Build Platform / OS Linux - linuxmint 21.3
And in Firefox 146.0.1 on the same machine the URL doesn't get updated.
And if you are open to bug reports.. if I move around the drawings move smoothly with the map, but if I zoom in/out the drawings move only after the map zooming animation ends, rather than smoothly
Could we also add text annotations? Also the delete button could delete just the last shape or a selected shape so as not to start over?
> It is RECOMMENDED that all senders and recipients support, at a minimum, URIs with lengths of 8000 octets in protocol elements. Note that this implies some structures and on-wire representations (for example, the request line in HTTP/1.1) will necessarily be larger in some cases.
Mainstream browsers support at least 64,000 characters [1], and Chrome supports up to 2MB [2].
[0]: https://www.rfc-editor.org/rfc/rfc9110#section-4.1-5
[1]: https://stackoverflow.com/a/417184/
[2]: https://chromium.googlesource.com/chromium/src/+/HEAD/docs/s...
Here is the Crime and Punishment by Fyodor Dostoevsky:
- https://medv.io/goto/crime-and-punishment-by-fyodor-dostoevs...
https://github.com/swiftlang/swift-corelibs-foundation/blob/...
My absolute favorite thing about modernity is how enabled we are to riff on a riff of a riff.
In 1346, if a blacksmith came up with something cool, its quite possible that it died with them.
Edit: actually not true since you use a url shortener
EDIT: actually I can edit the URL, but it takes a while to load.
It's truly insane how large typical share-URLS for content on instagram, youtube or any other large platforms are. URLs that could've been example.com/t/some-large-enough-id?time=13337 are stuffed with hundreds of characters, just to gather more data on people using these links.
> It is RECOMMENDED that all senders and recipients support, at a minimum, URIs with lengths of 8000 octets in protocol elements.
It is always worth remembering that, unless you have already ensured that the content has been rendered into a URI-safe subset of ASCII, a character and an octet are not the same thing.
https://developer.mozilla.org/en-US/docs/Web/JavaScript/Refe...
https://gabrielsroka.github.io/webpages/calc.htm#a1:=Rate=3....
More examples https://gabrielsroka.github.io/webpages/It's about 130 js loc
They can't track every website with the link and ask to be removed, either.
Could they ask textarea.my to not parse the link and thus, not display the content? Could textarea.my refuse?
Your example sounds like stopping notepad from rendering copyrighted content
From a regulatory perspective, it seems unlikely that most courts would appreciate the difference. In their mind - you run a website, and that website contains copyrighted content. Take it down.
You'd probably have to just blacklist the link in question to avoid a legal headache.
Edit. Call me a hater, but... I know the guy! That's the guy from Google whose code never works in the most hilarious ways! See issues on the rest of his pinned repos.
https://htmlpreview.github.io/?https://raw.githubusercontent...
Aand im dropped back to empty editor with just that one character visible
(Firefox 146.0.1 (Build #2016132551), 86bb7f6af6312ba3c0161085f854bcdff68f1a91 GV: 146.0.1-20251217121356 AS: 146.0.2 OS: Android 14)
https://github.com/planbnet/guitartabs
The Crime and Punishment one consistently crashes Brave mobile for me. I assume it's the length of the URL - and seen another commentator say the same for chrome mobile (sure they both use the same codebase so likely an upstream issue).
https://gist.github.com/smcllns/8b727361ce4cf55cbc017faaefbb...
data:text/html,<title>Notepad</title><textarea autofocus spellcheck=0 style="position:fixed;inset:0;padding:1em;border:0;font:monospace">
Your text actually survives a reboot in Chrome.
Can anyone think of a way to store the textarea value in the URL? I tried using JS to set a # but it's nonsensical in this context.
Edit: here's the best I could do:
data:text/html,<title>Notepad</title><textarea id=t autofocus spellcheck=0 style=position:fixed;inset:0;padding:2em;border:0;font:monospace></textarea><a id=s style=position:fixed;top:10px;right:10px>Right-click Open to save...</a><script>[,P,S]=location.href.slice(15).match(/(.<textarea[^>]>)[^]?(<\/textarea>.)/),t.oninput=U=_=>s.href='data:text/html,'+P+encodeURIComponent(t.value.replace(/&/g,'&').replace(/<\/textarea/g,'</textarea'))+S,U()</script>
I don't think urls were built for that kind of punishment.
It also has a note/plain text sharing option.
- https://textarea.my/#TYuxDcIwEEWpmeKUCiSIJQoKU0KFRBUWOGwnWDi...
My focus was on finding a good text→URL-slug compression strategy. I used ChatGPT-5.2-Pro mainly to explore and compare different compression approaches and trade-offs.
If you click save you get the option to use a URL.
The problem with a URL every edit is a new URL. So you send the URL to a friend, then fix a typo, they need a new URL.
The other problem is of course the space limit.
I built Ponder in the same vein. It, however, has 10 files. I did not use the URL, did not have double the fun, and now I’m sad.
https://github.com/codazoda/ponder
For fun I put it in chatgpt and asked if there are bugs.
It warned about fromBase64() and toBase64() not existing in main browsers. It is supported but is indeed a new "baseline 2025"feature. It suggested more compatible code using two small functions to convert characters manually.
"deflate-raw is not consistently supported." It suggested using 'deflate' instead.
The sharing works just like here, by encoding the tab itself in the url.
https://github.com/planbnet/guitartabs
you may think safari has no effective url limit (i.e. very high) but if you ever treat a url within the url bar as editable you are at risk to be silently truncated to 4096 bytes (eg select a character in the url bar and replace it)
also re-testing potential ~buffer limits in various ways on ipadOS 26.2 safari just now slowed my safari ui down to a crawl
eg after saving example.com with ~20k #hashtag to reading list -- each keystroke in this reply was taking several seconds, so I had to force quit safari and retype to post this warning
Students are lazy, in a good way, so they are more likely to run things on their own and play with interactive bits if the whole lecture is just one link.
Not really… using js to change the CSS on the go is not a good practice. Why does it matter? Because of the “dark mode” browser extensions. They often use the presence of @media query (or other standard CSS means of setting dark mode colors), and if it’s the JS that changes the colors we often get partial Dark Mode, which does not work at all.
Now if you bootstrap the app code into the url too then you can have a minimal kernel to run any machine in url.
Then you can also make a Quine somehow.
hopefully mine can stand out with all the extra features i have managed to cram in
I also have no way to confirm that URLs aren't logged server side, so I'd never trust the claim about "no tracking". That's why these projects also end up self-hosted.
Percent encoding is particularly bad since it may also bloat the length causing truncation and the decompress to fail. There's endless footguns with URLs.
[1]: https://davidlowryduda.com/mathshare/
--edit-- test link: https://linqshare.com/#eJxtkM9KxDAQxl-lzLmHrv8Ova3IHlz04BY8F...
https://gourav.io/devtools/notepad
data:text/html, <html contenteditable>
https://sqlscope.netlify.app/
Safari 15.6.1: Unhandled Promise Rejection: ReferenceError: Can't find variable: CompressionStream
but when I hit the keyboard I can see my it's is already loaded
Good job!
Firefox seems to work.
[0]: http://about.bitty.site/
Half a megabyte for a URL. That certainly is a thing.
https://flems.io/
https://textarea.my/#7cGBAAAAAMMgzfmTHORVAQAAAAAAAADAuwE=