aeyes

That could be caused by EU regulation because they have to meet fleet emission standards. Pretty much the only way to get there is to sell some EVs.

You don't need an EV charger. I charge my car from a normal wall socket over night, limited to 12A. That gives me 60 miles of range over night.

On Windows its not the same.

- MDI applications, for example old Photoshop and old Excel, can have multiple child windows inside of a single window. If you close the main window with the X button then everything closes.

- If you start multiple instances of an application you'll see multiple instances in the taskbar. On MacOS this is not really the intended way, you are supposed to have multiple windows. Afaik you can only launch multiple instances from the command line.

On Windows there are apps that minimize to tray instead of closing. I'd say that this is a similar problem but its a bit better than Slack on MacOS that just minimizes to Dock when you click the X button.

> You might want to read Apple human interface guidelines for what those buttons actually do.

Almost every app on MacOS does whatever they want with the minimize and close buttons. They are completely useless to me, you never know what will happen. And this isn't a recent problem, the way multiple windows are treated as one app is just unintuitive.

I don't think that nuclear projects of any size could pass current environmental impact assessments.

Only a severe energy shortage could motivate politicians enough to temporarily skip these, like they did for the LNG terminals. But you can't build a nuclear power plant of any size within a single legislative period so its just not going to happen.

The reason is money. They want regulation to make it harder for new competitors and competitors from other countries.

They invested billions into training the models but there is no competitive advantage, we see that within a couple of months everyone catches up. There is no way to profitability unless they get some policies to shields them against competitors that can't comply with the regulatory requirements.

That is also why there are things like Claude, Codex and Cursor. They are trying hard to build a customer relationship with a higher switching cost that hopefully sticks.

But the problem is that the AI buildout has become a large percentage of GDP. So obviously the government wants to keep it going because these companies are pumping enormous amounts of money into the economy.

The cold start numbers on that blog post show that its faster than AWS Lambda but your linked benchmark page [1] shows that Lambda is not only much faster but also much more consistent.

[1] https://cold.picheta.me/#bare

The author is not rebuilding anything, they just stuck a Raspberry Pi in the case.

My stream doesn't have constant pressure throughout the process ;-)

net income was 80 billion in the last 2 years alone, what are you talking about?

A patch for this was submitted in April: https://lists.ffmpeg.org/archives/list/ffmpeg-devel@ffmpeg.o...

Edit: And there was discussion about this back in 2024 as well

If those 1000 devs write to different branches then you only take a light lock on the branch, usually that is not the problem.

The problems start when there is a lot of changes on one branch and usually the first band-aid is to put a commit queue in place. The few very big companies that need even more than that probably aren't going to put their code on a new git hosting provider tomorrow.

Yes, they made it customizable. Here is a video showing the new options: https://www.youtube.com/watch?v=zX_WsBnuLf8

I was there before 2018, their reliability was never that good.

I still remember that they launched GitHub Actions in 2018 and back then I was thinking: They can't even make the website reliable, surely this will only cause them more problems.

The kernel is solid, its all the userspace software where they are fumbling hard.

For intelligence, I expect the next breakthrough to be colocation of memory and compute in the same chip. And we'll need much more of this memory, probably a few petabytes.

It's a terminal emulator with a command prompt that also gives you access to LLMs. But the interesting bit for people who work on a lot of machines is that this also works when you ssh to any remote machine.

So you can bring your agents to any remote system, it even works somewhat well for network devices.

And now there are tools like Warp Terminal which make me feel like all that knowledge about command line tools is just dead weight in my brain.

> Anthropic will probably go for $2T IPO now

How does that work? To me it sounds like there is an enormous risk that they get regulated to death. They can't sell their product.

It's not just you, I have the same problem.

They are popular because they are free.

I have to use Gmail at work and it is just terrible.

This data set was released years before the Covid hypergrowth phase which they are referring to.

> you should own that layer yourself inside of your infrastructure

Unless you have millions of users, you don't really need this. It would be nice to have but its not a pressing need. So why invest into developing something that you only need once you are at massive scale? At this point you might as well switch away from Postgres because you'll surely have the manpower to do it.

Even with a proxy like PgDog the Postgres sharding story isn't solved. Resharding with logical replication is unlikely to work with databases which are already TBs in size. I never got it to catch up, I had to sync data at the filesystem level which is terrible. Tools like pg_repack also fall apart at scale.

For those that get to a point where a sharding proxy is required, switching databases is a very appealing solution.

And for those that are almost there, application side sharding is more flexible than building a query routing proxy.

Where I work % of code written by AI and AI spend is tracked all the way down to the individual person.

It is obvious to me that this will be used in performance reviews in the future.

> Intel really made themselves unpopular with Apple during that period.

You can't tell me that this wasn't known by Apple before shipping the product. Why did they not provide adequate cooling for the CPU?

But you don't need a Dropbox account to view any file for which you created a shareable link.

If you add raw=1 to the URL then it will directly show in the browser without the Dropbox viewer.

Or did you share a folder?

It is running on a single machine so the player count is relatively low. Today the limit is ~1500 concurrent players but it used to be much lower. And of the players online probably half are bots which don't build.

The map is 30Mx30M blocks so outside of the very center you'll not see that much player activity.

DonutSMP is the largest Minecraft server today, the overworld is 225,000x225,000 blocks and the map is modified wherever you look.

LATAM as well, all major submarine cables land on the east coast. Surprisingly even from Mexico the latency is often better to US East.

Info from veeery long ago because I have been out of this stuff for over a decade:

The release will have an .sfv file with a CRC32 checksum for each rar file.

The FTP server checks them after the upload completes. Back in the day glftpd with zipscript was a very popular tool to manage an FTP site. This Readme sums it up well: https://github.com/pzs-ng/pzs-ng

The sfv can be tampered with but the propagation of releases to FTPs happens very fast, within minutes. It would take you longer to meaningfully alter it than it takes the racers to distribute the original files. And once the release is completely uploaded you can't modify the files anymore.

If the release is bad, for example if it doesn't work at all or if it contains a virus, then it simply gets nuked. This propagates within minutes.