Planet GNOME

There are a lot of prime classes, such as left truncating primes, twin primes, mersenne primes, palindromic primes, emirp primes and so on. The Wikipedia page on primes lists many more. Recently I got to thinking (as one is wont to do) how difficult would it be to come up with a brand new one. The only reliable way to know is to try it yourself.

The basic loop

The method I used was fairly straightforward:

1. Download a list of the first one million primes
2. Look at it
3. Try to come up with a pattern
4. Check if numbers from your pattern show up on OEIS
5. Find out they are not
6. Rejoice
7. Check again more rigorously
8. Realize they are in fact there in a slightly different form
9. Go to 2

Eventually I managed to come up with a prime category that is not in OEIS. Python code that generates them can be found in this repo. It may have bugs (I discovered several in the course of writing this post). The data below has not been independently validated.

Faro primes

In magic terminology, a Faro shuffle is one that cuts a deck of cards in half and then interleaves the results. It is also known as a perfect shuffle. There are two different types of Faro shuffle, an in shuffle and an out shuffle. They have the peculiar property that if you keep repeating the same operation, eventually the deck returns to the original order.

A prime p is a Faro prime if all numbers obtained by applying Faro shuffles (either in or out shuffles, but only one type) to its decimal representation are also prime. A Faro prime can be an Faro in prime, a Faro out prime or both. As an example, 19 is a Faro in prime, because a single in shuffle returns it to its original form. It is not an Faro out prime, because out shuffling it produces 91, which is not a prime (91 = 7*13).

The testing for this was not rigorous, but at least OEIS does not recognize it.

Statistics

I only used primes with an even number of digits. For odd number of digits you'd first need to decide how in and out shuffles should work. This is left as an exercise to the reader.

Within the first one milllion primes, there are 7492 in primes, 775 out primes and 38 that are both in and out primes.

The numbers with one or two digits are not particularly interesting. The first "actual" Faro in prime is 1103. It can be in shuffled once yielding 1013.

For the first out shuffle you need to go to 111533, which shuffles to 513131 and 153113.

The first prime longer than 2 digits that qualifies for both a Faro in and out prime is 151673. Its in shuffle primes are 165713, 176153 and 117563. The corresponding out shuffle primes are 151673, 617531 and 563117.

Within the first one million primes the largest in shuffle prime is 15484627, the largest out shuffle prime is 11911111 and the largest in and out prime is 987793.

Further questions

As is typical in maths, finding out something immediately raises more questions. For example:

Why are there so many fewer out primes than in primes?

How would this look for primes with odd number of digits in them?

Is it possible to build primes by a mixture of in and out shuffles?

Most of the primes do not complete a "full shuffle", that is, they repeat faster than a deck of fully unique playing cards would. For any number n can you find a Faro prime that requires that many shuffles or is there an upper limit for the number of shuffles?

In my last post on this  topic, I explained the history of SVG in GTK, and how I tricked myself into working on an SVG renderer in 2025.

Now we are in 2026, and on the verge of the GTK 4.22 release. A good time to review how far we’ve come.

Testsuites

While working on this over the last year, I was constantly looking for good tests to check my renderer against.

Eventually, I found the resvg testsuite, which has broad coverage and is refreshingly easy to work with. In my unscientific self-evaluation, GtkSvg passes 1250 of the 1616 tests in this testsuite now, which puts GTK one tier below where the web browsers are. It would be nice to catch up with them, but that will require closing some gaps in our rendering infrastructure to support more complex filters.

The resvg testsuite only covers static SVG.

Another testsuite that I’ve used a lot is the much older SVG 1.1 testsuite, which covers SVG animation. GtkSvg passes most of these tests as well, which I am happy about — animation was one of my motivations when going into this work.

https://blogs.gnome.org/gtk/files/2026/02/Screencast-From-2026-02-24-20-45-33.webm Benchmarks

But doing a perfect job of rendering complex SVG doesn’t do us much good if it slows GTK applications down too much. Recently, we’ve started to look at the performance implications of SVG rendering.

We have a ‘scrolling wall of icons’ benchmark in our gtk4-demo app, which naturally is good place to test the performance impact of icon rendering changes. When switching it over to GtkSvg, it initially dropped from 60fps to around 40 on my laptop. We’ve since done some optimizations and regained most of the lost fps.

The performance impact on typical applications will be much smaller, since they don’t usually present walls of icons in their UI.

Stressing our rendering infrastructure with some more demanding content was another motivation when I started to work on SVG, so I think I can declare success here.

Content Creators

The new SVG renderer needs new SVGs to take advantage of the new capabilities. Thankfully, Jakub Steiner has been hard at work to update many of the symbolic icons in GNOME.

Others are exploring what we can do with the animation capabilities of the new renderer. Expect these things to start showing up in apps over the next cycle.

Future work

Feature-wise, GtkSvg is more than good enough for all our icon rendering needs, so making it cover more obscure SVG features may not be big priority in the short term.

GtkSvg will be available in GTK 4.22, but we will not use it for every SVG icon yet — we still have a much simpler symbolic icon parser which is used for icons that are looked up by icon name from an icontheme. Switching over to using GtkSvg for everything is on the agenda for the next development cycle, after we’ve convinced ourselves that we can do this without adverse effects on performance or resource consumption of apps.

Ongoing improvements of our rendering infrastructure will help ensure that that is the case.

Where you can help

One of the most useful contributions is feedback on what does or doesn’t work, so please: try out GtkSvg, and tell us if you find SVGs that are rendered badly or with poor performance!

Update: GtkSvg is an unsandboxed, in-process SVG parser written in C, so we don’t recommend using it for untrusted content — it is meant for trusted content such as icons, logos and other application resources. If you want to load a random SVG of unknown providence, please use a proper image loading framework like glycin (but still, tell us if you find SVGs that crash GtkSvg).

Of course, contributions to GtkSvg itself are more than welcome too. Here is a list of possible things to work on.

If you are interested in working on an application, the simple icon editor that ships with GTK really needs to be moved to its own project and under separate maintainership. If that sounds appealing to you, please get in touch.

If you would like to support the GNOME foundation, who’s infrastructure and hosting GTK relies on, please donate.

Its moments of change that remain striking in your memory when you look back. I feel like i’m in a long period of change, and if like me you participate in the tech industry and open source then you probably feel the same. It’s going to be a wild time to look back on.

As humans we’re naturally drawn to exciting new changes. Its not just the tech industry. The Spanish transport minister recently announced ambicious plans to run trains at record speeds of 350km/h. Then two tragic accidents happened, apparently due to careless infrastructure maintenance. Its easy (and valid) to criticise the situation. But I can sympathise too. You don’t see many news reports saying “Infrastructure is being maintained really well at the moment and there haven’t been any accidents for years”. We all just take that shit for granted.

This is a “middle aged man states obvious truths” post, so here’s another one we forget in the software world: Automating work doesn’t make it go away. Lets say you automate a 10 step release process which takes an hour to do manually. That’s pretty great, now at release time you just push a button and wait. Maybe you can get on with some other work meanwhile — except you still need to check the automation finished and the release published correctly. What if step 5 fails? Now you have drop your other work again, push that out of your brain and try to remember how the release process worked, which will be hazy enough if you’ve stopped ever doing release manually.

Sometimes I’ll take an hour of manual work each month in preference to maintaining a complex, bespoke automation system.

Over time we do build great tools and successfully automate bits of our jobs. Forty or fifty years ago, most computer programmers could write assembly code and do register allocation in their heads. I can’t remember the last time I needed that skill. The C compiler does it for me.

The work of CPU register allocation hasn’t gone away, though. I’ve outsourced the cognitive load to researchers and compiler teams working at places like IBM / Red Hat, embecosm and Apple who maintain GCC and LLM.

When I first got into computer programming, at the tail end of the “MOV AX, 10h; INT 13h” era, part of the fun was this idea you could have wild ideas and simply create yourself piece by piece, making your own tools, and pulling yourself up by your bootstraps. Look at this teenager who created his own 3D game engine! Look at this crazy dude who made an entire operating system! Now I’m gonna do something cool that will change the world, and then ideally retire.

It took me the longest time to see that this “rock star” development model is all mythology. Just like actual rock stars, in fact. When a musician appears with stylish clothes and a bunch of great songs, the “origin story” is a carefully curated myth. The music world is a diverse community of artists, stylists, mentors, coaches, co-writers, producers, technicians, drivers, promotors, photographers, session musicians and social media experts, constantly trading our skills and ideas and collaborating to make them a reality. Nobody just walks out of their bedroom onto a stage and changes the world. But that doesn’t make for a good press release does it ?

The AI bubble is built on this same myth of the individual creator. I think LLMs are a transformative tool, and computer programming will never be the same; the first time you input some vaguely worded English prompt and get back a working unit test, you see a shining road ahead paved with automation, where you can finally turn ideas into products within days or weeks instead of having to chisel away at them painfully for years.

But here’s the reality: our monkey brains are still the same size, and you can’t If your new automation is flaky, then you’re going to spend as much time debugging and fixing things as you always did. Doing things the old way may take longer, but the limiting factor was never our typing speed, but our capacity to understand and communicate new ideas.
“The future belongs to idea guys who can just do things”. No it doesnt mate, the past, present and future belongs to diverse groups of people whose skills and abilities complement each other and who have collectively agreed on some sort of common goal. But that idea doesn’t sell very well.

If and when we do land on genuinely transformative new tool — something like a C compiler, or hypertext — then I promise you, everyone’s going to be on it in no time. How long did it take for ChatGPT to go from 0 to 1 billlion users wordwide?

In all of this, I’ve had an intense few months in a new role at Codethink. It’s been an intense winter too — by some measures Galicia is literally the wettest place on earth right now — so I guess it was a good time to learn new things. Since I rejoined back in 2021 I’ve nearly always been outsourced on different client projects. What I’m learning now is how the company’s R&D division works.

You all know that librsvg is developed in gitlab.gnome.org, not in GitHub. The README prominently says, "PLEASE DO NOT SEND PULL REQUESTS TO GITHUB".

So, of course, today librsvg got its first AI slop pull request and later a second one, both in GitHub. Fortunately (?) they were closed by the same account that opened them, four minutes and one minute after opening them, respectively.

I looked.

There is compiled Python code (nope, that's how you get another xz attack).

There are uncomfortably large Python scripts with jewels like subprocess.run("a single formatted string") (nope, learn to call commands correctly).

There are two vast JSON files with "suggestions" for branches to make changes to the code, with jewels like:

• Suggestions to call standard library functions that do not even exist. The proposed code does not even use the nonexistent standard library function.

• Adding enum variants to SVG-specific constructs for things that are not in the SVG spec.

• Adding incorrect "safety checks". assert!(!c_string.is_null()) to be replaced by if c_string.is_null() { return ""; }.

• Fix a "floating-point overflow"... which is already handled correctly, and with a suggestion to use a function that does not exist.

• Adding a cache for something that does not need caching (without an eviction policy (so it is a memory leak)).

• Parallelizing the entire rendering process through a 4-line function. Of course this does not work.

• Adding two "missing" filters from the SVG spec (they are already implemented), and the implementation is todo!().

It's all like that. I stopped looking, and reported both PRs for spam.