I think stock market crashes don't have to be so bad. We have now seen a long period of companies getting more and more powerful and caring less and less about the customers. Luckily, a stock market crash hurts public companies and people who own a lot of stocks the hardest. I mean I wish that my net worth would have gone down by 1 billion last week, but it didn't. I didn't have 1 billion to begin with so well not much happened.

Having said that, yes recessions can be tough. Some (not all!) people do lose their job. But sometimes it is necessary to cleanup.

Luckily for you, the recession could be great opportunity for you. You are young and nimble. If the world changes, you can adapt. The same cannot be said from old companies. They are not nimble and some will disappear. That's how things go.

In tegenstelling tot de meeste andere hier zal ik dan maar zeggen wat je eigenlijk al weet: mensen doodwensen is waarschijnlijk niet de oplossing. In mijn geval ben ik tegen de doodstraf dus ook tegen mensen doodwensen.

Als extra argument, het is mentaal ook helemaal niks aan om mensen dood te wensen. Het levert alleen maar een negatieve mindset op. Volgens mij veel leuker om het leven te focussen op goede dingen op te bouwen in plaats van slechte dingen af te breken.

I think China is not only promoting the (extreme) left as you suggest but rather the (extreme) right too.

It's the age old "divide and conquer" strategy (https://en.m.wikipedia.org/wiki/Divide_and_conquer).

Anyone who makes money from oil is also probably piling in.

and ai should only be used as a copilot not as a pilot.

I use Cursor a lot but 99% of the time in copilot mode. Mostly just small suggestions and refactorings, and sometimes I select bigger areas and then CTRL + K and I ask it what should change about it.

It sounds poetic when written as "The thing worse than a bully is a bully that pretends to not be one."

The problem is that you replaced deficiencies of GH actions with deficiencies of Rust cargo.

Yes thanks also for your comment. I fully agree and I mention in the blog that it's not perfect. Probably I should try to package the tool into Debian packages. Or do you know another delivery method that is easy to set up but still safe? I wish Nix was available but alas https://github.com/actions/runner-images/issues/1579.

And my humble attempt at making cargo a little bit more secure https://www.reddit.com/r/rust/comments/1d6zs8s/cargo_and_supply_chain_attacks/

Great point in this one! I agree that sandboxing during build would be a great security measure.

What’s wrong with installing ffmpeg from apt?

Yes security-wise apt is fine indeed. Benefits of installing the binaries could be that you know exactly which binary you are running. I think this reproducibility can be very useful especially in GitHub workflows since dependencies that silently change can be very hard to debug. It also is a bit faster (25 seconds vs. 10 seconds).

Also where it installs the package? Does it add binaries to the PATH?

By default in ~/.jas/bin/. This can be modified by setting --dir.

The notion that increasing the money supply increases inflation has not been true in the US, UK and Japan for decades, probably due to modern manufacturing and farming.

What is the point of the FED then? Why can't the FED lower interest rates due to "high inflation" (e.g., https://www.wsj.com/economy/cpi-inflation-january-2025-interest-rate-34aa95db).

The inflation was caused by money printing. I don't see how tax cuts cause inflation.

EDIT: For clarification, the narrative around the Fed is that they basically look at inflation and then determine the interest rate. See for example https://www.wsj.com/economy/cpi-inflation-january-2025-interest-rate-34aa95db.

Not always. In financial crises sometimes the cheaper products like lipstick and toys will still be bought but more expensive items like a new car or a new house not. It’s called the lipstick effect.

Thanks for the extra info very interesting!

Very good point on the dell.

On the second point I still somewhat disagree, but I'll think about it and maybe change my mind. I've read some benchmarks were those intel chips in that generation were even throttled in more serious desktop setups (heat generation increases exponential at some point with clock speed if I remember correctly), but I'll think about it

If you're interested, I'll elaborate:

100%.

I agree with most of what you said. One small nuance though that I think is overlooked: in my experience, the beter thermals lead to better performance. The Intel chips might look fine on benchmarks but will fail in the real world. I used to have a laptop that would grind to a halt after 20 minutes of compiling already. The system would just start freezing. So if the benchmark is 5 minutes it might look like the Intel is doing great, but in practise they are not. With the Apple benchmarks, they can achieve the high scores consistently, in my experience. I'm currently running an M2 Pro and it is nearly impossible to get the fan spinning. For comparison, a $1000 Dell with an Intel chip 2 years ago (I bought it new back then) would take 40 minutes to compile LLVM, while the system was unusable during compilation. When I went to an M1, the system remained usuable and compiled in 20 minutes. My current M2 Pro at 10 minutes (that's also why I bought it). The newest M4 mini should also be able to do it in 10 minutes if I can believe the benchmarks.

Great argument you really convinced me there

The $800 mac mini is?!

Yes GPUs easily. GPUs and CPUs like the Apple M-series execute way more computations per cycle than before. Also, memory moved closer and many other hardware tweaks. Just look at the Apple M4 mini benchmarks. That thing is insane for the price. Same if you compare GPU inference speed with modern FP8 or even lower to the old 64 or 32-bit. 

Gym machines, for example, are well-supported in the literature for hypertrophy and have similar or better results as free weights.

Although I have not read Taleb's arguments on gym machines, I personally don't like gym machines anymore either. The counterargument is often that loose weights require you to use more small muscles, which means you get a lot more "real world" useful strength. Also it just feels better to have free weights if you ask me. Less chance of having a movement be too constrained and therefore having to move in weird angles.

Having said that, I agree that Taleb sometimes doesn't really know what he is talking about. But that's okay I guess. I'll just ignore it usually.

Oh yes I see only when things are published they are true

  It doesn't seem that there's any evidence at scale that LLMs/chatbots are providing more benefit than problems.

What world do you live in? I program for 10 years and my speed went up by 20-40%. I mainly have to type less because the LLM can guess what I want to type in many cases. It also comes with bad suggestions but those I just ignore.

And I honestly don’t know what is now the problem with low quality bug reports and PRs. This has been a “problem” for many years already. Most maintainers are very skilled in judging whether they will spend time on an issue or PR or not. The unfortunate reality is that most of them will go stale. Very sad for people (like me from time to time) who have put in real effort into the PR, but that’s life.

If you and your partner find it important to marry then please marry. It's 100% more important than your career. It sounds harsh, but the reality is that your career doesn't care about you.

Canceling talks not normal? Countless talks have been cancelled in the last 10 years. Imagine for example being a professor opposed to covid vaccines. Many of them got cancelled.

Unless many people in the faculty are historians, I would say the combined experience is only 50 or so years. Everyone lived the same era so mostly the same experiences. Every time I listen to a historian I realize how little we know from the past. 

Not really. By default it has some very weird default behavior for many of these utils. You can improve the situation but I never found the Apple utils very enjoyable. That's also why I mentioned many modern Rust tools in another comment. It's way easier to run cargo install <tool> than randomly hitting a MacOS tool difference.

You mean fd, rg, bat, hyperfine, and just?

My pen just fell from the table. Probably Elon's fault too.