How to get banned from Facebook in one simple step

I, too, have (or as you can probably guess from the title of this post, had) a Facebook account. I only ever used it for two purposes.

1. Finding out what friends I rarely see are doing
2. Getting invites to events

Facebook has over the years made usage #1 pretty much impossible. My feed contains approximately 1% posts by my friends and 99% ads for image meme "humor" groups whose expected amusement value seems to be approximately the same as punching yourself in the groin.

Still, every now and then I get a glimpse of a post by the people I actively chose to follow. Specifically a friend was pondering about the behaviour of people who do happy birthday posts on profiles of deceased people. Like, if you have not kept up with someone enough to know that they are dead, why would you feel the need to post congratulations on their profile pages.

I wrote a reply which is replicated below. It is not accurate as it is a translation and I no longer have access to the original post.

Some of these might come via recommendations by AI assistants. Maybe in the future AI bots from people who themselves are dead carry on posting birthday congratulations on profiles of other dead people. A sort of a social media for the deceased, if you will.

Roughly one minute later my account was suspended. Let that be a lesson to you all. Do not mention the Dead Internet Theory, for doing so threatens Facebook's ad revenue and is thus taboo. (A more probable explanation is that using the word "death" is prohibited by itself regardless of context, leading to idiotic phrasing in the style of "Person X was born on [date] and d!ed [other date]" that you see all over IG, FB and YT nowadays.)

Apparently to reactivate the account I would need to prove that "[I am] a human being". That might be a tall order given that there are days when I doubt that myself.

The reactivation service is designed in the usual deceptive way where it does not tell you all the things you need to do in advance. Instead it bounces you from one task to another in the hopes that sunk cost fallacy makes you submit to ever more egregious demands. I got out when they demanded a full video selfie where I look around different directions. You can make up your own theories as to why Meta, a known advocate for generative AI and all that garbage, would want a high resolution scans of people's faces. I mean, surely they would not use it for AI training without paying a single cent for usage rights to the original model. Right? Right?

The suspension email ends with this ultimatum.

If you think we suspended your account by mistake, you have 180 days to appeal our decision. If you miss this deadline your account will be permanently disabled.

Well, mr Zuckerberg, my response is the following:

Close it! Delete it! Burn it down to the ground! I'd do it myself this very moment, but I can't delete the account without reactivating it first.

Let it also be noted that this post is a much better way of proving that I am a human being than some video selfie thing that could be trivially faked with genAI.

AI and money

If you ask people why they are using AI (or want other people to use it) you get a ton of different answers. Typically none of them contain the real reason, which is that using AI is dirt cheap. Between paying a fair amount to get something done and paying very little to give off an impression that the work has been done, the latter tends to win.

The reason AI is so cheap is that it is being paid by investors. And the one thing we know for certain about those kinds of people is that they expect to get their money back. Multiple times over. This might get done by selling the system to a bigger fool before it collapses, but eventually someone will have to earn that money back from actual customers (or from government bailouts, i.e. tax payers).

I'm not an economist and took a grand total of one economics class in the university, most of which I have forgotten. Still, using just that knowledge we can get a rough estimate of the money flows involved. For simplicity let's bundle all AI companies to a single entity and assume a business model based on flat monthly fees.

The total investment

A number that has been floated around is that AI companies have invested approximately one trillion (one thousand billion or 1e12) dollars. Let's use that as the base investment we want to recover.

Number of customers

Sticking with round figures, let's assume that AI usage becomes ubiquitous and that there are one billion monthly subscribers. For comparison the estimated number of current Netflix subscribers is 300 million.

Income and expenses

This one is really hard to estimate. What seems to be the case is that current monthly fees are not enough to even pay back the electricity costs of providing the service. But let's again be generous and assume that some sort of a efficiency breakthrough happens in the future and that the monthly fee is $20 with expenses being $10. This means a $10 profit per user per month.

We ignore one-off costs such as buying several data centers' worth of GPUs every few years to replace the old ones.

The simple computation

With these figures you get $10 billion per month or $120 billion per year. Thus paying off the investment would take a bit more than 8 years. I don't personally know any venture capitalists, but based on random guessing this might fall in the "takes too long, but just about tolerable" level of delay.

So all good then?

Not so fast!

One thing to keep in mind when doing investment payback calculations is the time value of money. Money you get in "the future" is not as valuable as money you have right now. Thus we need to discount them to current value.

Interest rate

I have no idea what a reasonable discount rate for this would be. So let's pick a round number of 5.

The "real-er" numbers

At this point the computations become complex enough that you need to break out the big guns. Yes, spreadsheets.

Here we see that it actually takes 12 years to earn back the investment. Doubling the investment to two trillion would take 36 years. That is a fair bit of time for someone else to create a different system that performs maybe 70% as well but which costs a fraction of the old systems to get running and operate. By which time they can drive the price so low that established players can't even earn their operating expenses let alone pay back the original investment. 

Exercises for the reader

• This computation assumes the system to have one billion subscribers from day one. How much longer does it take to recuperate the investment if it takes 5 years to reach that many subscribers? What about 10 years?
• How long is the payback period if you have a mere 500 million paid subscribers?
• Your boss is concerned about the long payback period and wants to shorten it by increasing the monthly fee. Estimate how many people would stop using the service and its effect on the payback time if the fee is raised from $20 to $50. How about $100? Or $1000?
• What happens when the ad revenue you can obtain by dumping tons of AI slop on the Internet falls below the cost of producing said slop?

Converting Chapterizer from Cairo + Pango to CapyPDF

Chapterizer (not a great name, I know) is a tool I wrote to generate books. Originally used Cairo and Pango to generate PDF files. It works and was fairly easy to get started but has its own set of downsides:

• Cairo always produces RGB PDFs, which are not accepted by printing houses
• Cairo does not handle advanced PDF features like trim boxes
• Pango aligns text at the top of each line, but for high quality text output you have to do baseline alignment
• Pango is designed to "always print something", which is to say it does transparent font substitution for example when the chosen font does not have some glyph

I have also created CapyPDF to generate "proper" PDF. Over the holidays I finalized porting Chapterizer to use CapyPDF. The pipeline is now surprisingly simple. First you read in the source text, then it is shaped with Harfbuzz and then written to a PDF file with CapyPDF.

It was grunt work. Nothing about it was particularly difficult, just dealing with the same old issues like the fact that in PDF the page's origin is at bottom left, whereas in Cairo it is at the top left.

Anyhow, now that it is done we can actually test the performance of CapyPDF with a somewhat realistic setup. Currently creating a 40 page document takes 0.4 seconds which comes down to 0.01 seconds per page. Which is fast enough for me.

New year, new Pystd epoch, or evolving an API without breaking it

One of the core design points of Pystd has been that it maintains perfect API and ABI stability while also making it possible to improve the code in arbitrary ways. To see how that can be achieved, let's look at what creating a new "year epoch" looks like. It's quite simple. First you run this script

Then you add the new files to Meson build targets (I was too lazy to implement that in the script). Done. For extra points there is also a new test that mixes types of pystd2025 and pystd2026 just to verify that things work.

As everything is inside a yearly namespace (and macros have the corresponding prefix) the symbols do not clash with each other.

At this point in time pystd2025 is frozen so old apps (of which there are, to be honest, approximately zero) keep working forever. It won't get any new features, only bug fixes. Pystd2026, on the other hand, is free to make any changes it pleases as it has zero backwards compatibility guarantees.

Isn't code duplication terribly slow and inefficient?

It can be. Rather than handwaving about it, lets measure. I used my desktop computer which has an AMD Ryzen 7 3700X.

Compiling Pystd from scratch and running the test suite (with code for both 2025 and 2026) in both debug and optimized modes takes 3 seconds in total (1s for debug, 2s for optimized). This amounts to 2*13 compiler invocations, 2 static linker invocations and 2*5 dynamic linker invocations.

Compiling a helloworld with standard C++ using -O2 -g also takes 3 seconds. This amounts to a single compiler invocation.