Jorvik Software — Notes (Unofficial) / Jorvik Software

Unofficial automated RSS/Atom links from the source Jorvik Software — Notes, covering Samples include software release changelogs and an independent screenshot-renaming app. Original creators are credited at the source; follow original links. Zaps support the independent feed operator, not the source creators.

Last Notes

2026-10-20 01:49:14 GMT

Who Finds the Bugs?
Photo by Chris Benson on Unsplash
In April I wrote that my apps don’t phone home. No analytics, no crash reports, no usage data. In the same post I admitted what that costs: when someone hits a bug, I rely on them telling me, and most people won’t.
Six months on, there is an obvious question. If nothing reports back, who finds the bugs?
Mostly, I do. I am the telemetry.
An ordinary Monday
Today was a normal working day, and three of my apps let me down in it.
Jorvik Daily News is the first thing I read each morning. This afternoon I removed the lead story, and the next headline moved up into its place. The picture didn’t. For about half a minute, a headline about the internet sat under a drawing of New York’s dogs. The app had kept the old picture on screen until the new one finished downloading, and that site was slow.
A little later, the paper led with a portrait of a roboticist standing beside his robot. Both of them had lost their heads. The cropping code was wrong. Six weeks, and I never noticed.
Then I took a screenshot of a work conversation in Microsoft Teams, and Nomen named it after Ghostty, my terminal. macOS saves a screenshot a few seconds after you take it. In those few seconds I had switched to Ghostty to paste it, and Nomen asked which app was in front too late.
All three are fixed, and the fixes shipped this evening. None of them came from a dashboard or a crash log. Each one got in my way while I was trying to do something else.
Why it works
This only works because I am a real user, not a tester. First and foremost, I write these apps to satisfy my own needs. Nobody commissioned them. Each one exists because something on my Mac annoyed me enough to fix it.
So I read the paper every day because I want the news. Nomen renames my screenshots because I take dozens of them at work. I don’t open these apps to look for bugs. I open them to get something done, and a bug has to interrupt that before I see it.
That turns out to be a good filter. The bugs I find are the ones that cost a real person time. Nobody has to argue for them in a planning meeting. I have already felt them.
It helps that I am my own biggest critic. A picture cropped through someone’s forehead is not a bug report I file and come back to next quarter. It bothers me until it is fixed.
Why it isn’t enough
The robot shows the limit.
That crop bug was in every copy of the app for six weeks. It affected only the pictures that needed cropping, and on a typical front page that is about three in twenty-four. A centred crop often looks fine. So I read the paper every morning for six weeks and saw nothing wrong, until a picture arrived with the faces in exactly the wrong place.
One person sees what one person’s screen shows. My Mac, my displays, my feeds, my habits. The Nomen bug existed because I paste screenshots into a terminal. Somebody else’s habits would uncover bugs mine never will, and mine will keep hiding them.
The telemetry I didn’t collect
This is where other people come in.
When strangers started opening issues on GitHub, they did what telemetry is supposed to do. fuzzy76 found that MenuTidy ignored the first click on Tahoe. thechurchofthechad-dotcom told me that Jorvik Daily News had stopped opening articles at all. emjayess found that RainbowApple left its logo floating in mid-air when the menu bar hides itself. I don’t hide my menu bar. I would never have found that one.
Each of them told me something I could not have learned from my own usage. That is exactly what telemetry promises. The difference is that they chose to tell me. They decided what to send, they wrote it in their own words, and some sent screenshots and steps to reproduce.
Others sent something telemetry can’t send at all. RSS1102 translated QuitProtect into Simplified Chinese, and opened a pull request that exposed a significant bug I had shipped in several apps. Christophe has added six features to Save Cannes, from playback profiles for each display to an art mode for a single screen. None of that was a bug report. It was a person telling me what they needed, with the code to go with it. No dashboard has ever done that.
I wrote about that in August as a thank-you. It is also the other half of the decision I made in April. I gave up knowing things about my users in secret. What I get instead is a smaller number of people telling me things in the open, and I think that is the better deal.
Dog food
There is an old phrase for using your own product: eating your own dog food. The point is to suffer what your customers suffer, before they have to.
I do eat mine, every day, and today it was gritty in three places. But I am one dog. If something in one of these apps gets in your way, it may simply not have got in mine yet. The issue tracker for every app is open, and I read every report.
https://jorviksoftware.cc/notes/2026/10/19/who-finds-the-bugs
https://jorviksoftware.cc/notes/2026/10/19/puppy-bowl.1600w.jpg
#bugs #nomen #six #ghostty #mac #mondaytoday #newyork
#rss #feeds #linksharing #qwibbit

2026-10-17 03:52:14 GMT

Same Pictures, an Eighth of the Weight
Photo by Philip Oroni on Unsplash
This site has used JPEG and PNG images since the first page. I never thought about it. Then I read about AVIF, a newer image format, and I asked Claude if it was worth the change.
Claude answered with measurements. It converted some of my real images and compared the sizes. One screenshot from Strataris was 4.65 MB. As an AVIF, it was 67 KB.
So yeah, I just had to take this further.
Why so large?
AVIF was only half of it. Some of my header photos were 3,200 pixels wide. The column they sit in is 800 pixels wide, and a Retina screen needs 1,600 at most. I was effectively sending pixels nobody would see.
The build pipeline now fixes both. When I add a JPG or a PNG, the build makes a smaller copy. The copy is the largest size that the site ever draws the image. Then the build makes an AVIF of it. Icons get a smaller limit, because the site never draws one wider than 128 pixels.
The HTML offers both files. A browser that can read AVIF takes the AVIF. Safari has read AVIF since version 16, Chrome since 85 and Firefox since 93. An older browser takes the resized JPG or PNG.
I still add full-size originals, and I never touch the generated files. That part mattered most to me. Nothing about how I write a post had to change.
What it saved
Some single images:
Image
Before
After
A Strataris screenshot (PNG, 3,200 pixels wide)
4.65 MB
67 KB
The photo at the top of Typography Matters (JPG)
1.64 MB
89 KB
The Browser Notes icon (PNG, 1,024 pixels wide)
1.17 MB
10 KB
A Citadel banner (JPG, already small)
67 KB
15 KB
My photo on the About page
41 KB
16 KB
The Strataris post gained the most. It went from 12.39 MB to 196 KB. I had put three full-resolution PNG screenshots in it and never checked their size. That one is on me.
Citadel gained the least, because its screenshots were already sensible JPGs.
I was worried about the screenshots. Small text and thin lines show compression first, so PNG screenshots get a higher quality setting than photos. I compared them at full size, side by side. I could not tell them apart.
The two editors
I write posts in my Notes Editor, and I edit the rest of the site in my Web Editor. I built both of them. I expected to change both.
The Notes Editor needed nothing. It publishes by running the site’s own build script, so it got the AVIF step for free.
The Web Editor needed two small fixes. Its preview pane changes image addresses so that they load from the disk. It knew about src, but not about srcset, which the new <picture> tags use. So every converted image showed as broken in the preview. The editor also did not know that .avif is an image. It would have opened one as text, and a save would have destroyed it.
Then Claude found a larger problem. The Web Editor has a Visual mode, for editing a page as it looks. If I typed one word in it and saved, it wrote the browser’s version of the page over my file. The doctype and the whole <head> disappeared. Every … became a literal ellipsis.
A page that contained ${ opened empty, and one keystroke and a save put that empty page over the file. Visual mode also kept the old page when I opened a different file. A save then put that page into the new file.
I had never noticed.
Web Editor 1.0.26 fixes all of it. Visual mode now edits only the body of a page, and it writes back only the characters I change. Everything else keeps its exact bytes.
One more thing broke, on my staging server only. It copied .jpg files from the top folder of the site but not .avif files, so my photo on the About page was missing there. That took one line to fix.
https://jorviksoftware.cc/notes/2026/10/16/same-pictures-an-eighth-of-the-weight
https://jorviksoftware.cc/notes/2026/10/16/yellow-flower.1600w.jpg
https://jorviksoftware.cc/notes/2026/10/16/avif-image-weight.svg
#samepictures #avif #png #webeditor #weight #noteseditor #jpg #claude #strataris #browser
#rss #feeds #linksharing #qwibbit

2026-10-17 03:23:04 GMT

Seek, and Ye Shall Find
Photo by Emiliano Vittoriosi on Unsplash
Until this week, this site had no search. To find an old post, you paged through the Notes and hoped. That is fine for ten posts. There are more than seventy now.
So now there is a magnifying glass in the menu.
No server
I did not want a search server. This site is a set of static files on Cloudflare. A search server is one more thing to maintain, and a search service from a third party sees what my readers type. I did not want either.
Pagefind is an open-source tool from CloudCannon, and it works differently. It runs once, when I publish. It reads every page and writes an index, which is a folder of small files beside the site. When you search, your browser downloads only the parts of the index that your words need. Then your browser does the search.
The search engine is about 157 KB, and your browser downloads it once. Each search adds 30 to 80 KB. Indexing the whole site takes about 0.2 seconds.
What the index ignores
Every page has a menu and a footer. If Pagefind indexes them, every page matches the word “Donate”. So Pagefind ignores them.
Photo credits were the next problem. Each result shows a short extract from the page. On posts with a header photo, the extract was often “Photo by … on Unsplash”. True, but not useful. Pagefind ignores the credits now too.
The word that matched everything
I tested the search on my staging server. Then I typed “insane”. I got 111 results, which is every page in the index.
This is why. When Pagefind cannot find a word, it does not stop. It finds the longest indexed word that is the start of your word. “Insane” starts with “in”, and every page on this site contains “in”.
This is not carelessness. It is this behaviour that lets you search as you type: “spar” finds Sparkle before you finish the word. But it needs a limit, and other people have seen the same problem.
We found a way around it. Pagefind does not use the fallback for a word in quotation marks, but it still finds other forms of that word. So “building” in quotes still finds “build”. The search box now asks two questions about each word:
Does the word in quotes find any pages?
If not, do the words that Pagefind highlights start with what you typed? This keeps search as you type.
If both answers are no, Pagefind guessed. The box then shows “No results”, which is the correct answer. Now “insane” finds nothing, “spar” finds 45 pages and “building” finds 91.
One oddity stays. Search for “apps”, and the first result is Fine, Apple. You Win. Pagefind extends “app” to “apple”. That is the same feature that makes “spar” work, so it stays. I suspect Apple would approve of the ranking.
Our own search box
Pagefind includes a ready-made search box, but it cannot filter results. So we wrote a small one, 83 lines of JavaScript, on Pagefind’s API. It shows results as you type. It puts your search in the address field, so you can send a link to a search or bookmark one. It tells screen readers how many results there are.
command-K
I wanted a keyboard trigger. Claude first suggested the / key, as many sites use. But a shortcut that is one character fails WCAG 2.1.4. A person who controls the computer by voice could say a word that contains that character, and the search engine would activate. Unexpected behaviours like that are unforgivable.
So I chose commandK on a Mac and controlK on other computers. It opens the search page from any page. On the search page, it selects the search box. On a Mac, controlK does nothing new, because in a text field it deletes to the end of the line, and some people use that every day.
Try it. Press commandK and search for something. You never know what you might discover in the depths of this website.
https://jorviksoftware.cc/notes/2026/10/17/seek-and-ye-shall-find
https://jorviksoftware.cc/notes/2026/10/17/magnifying-glass.1600w.jpg
#rss #feeds #linksharing #qwibbit