0xCMP

ICYMI:

> Over the past several months, our team has rebuilt the infrastructure behind Edge Functions, working closely with the team at Unikraft, who wrote about the experience from their side.

The unfortunate part of that graph is that I think it's not updated for today's prices given the memory shortage/crunch we're experiencing that's driven up the prices for all kinds of memory.

But also while it should be faster than it is, how would an S3 designed around SSDs look differently to an API user? I would think the API is basically the same.

As someone with TBs of SSDs and HDDs laying around I am someone who agrees with your point, but this is not a good way to compare costs.

For one, this has no redundancy and doesn't factor the cost of the machine providing the storage. But also it's an upfront cost for storing 4TB. It would be implied you wouldn't literally store 4TB on the SSD because now you're out of capacity for more data. If you only have 2GB of data in a bucket then S3 is still magnitudes cheaper, faster, and more redundant than anything you could put together yourself because the cost is spread across all the users of the service.

Reality is there are so many times that S3 makes the most sense that it makes people short-circuit and always pick S3 despite the huge hidden IOPS and bandwidth costs everyone rightfully tries to point out.

This is great and I think for any company/individual who is going to ensure their domain is registered and maintained this makes a lot of sense. It would be catastrophic, but there is no reason Github couldn't disappear or otherwise change some policies that require moving away from it. The way Go works makes committing these package names tied to Github so much more weighty than simply the place you pull from.

The one thing that I was worried about was returning 301 in the example Nginx config. If you ever wanted to change the url that clients are redirected to then any browsers that visited the old url config would be forced to go to the old config's redirect url. For `go ...` and `curl` it wouldn't matter, but Chrome/Firefox will cache that 301 permanently and break the intended redirect. Not sure if this is really an issue in practice though.

> triggering a "de-fragmentation" of the available efivar space so that there's enough contiguous space to deploy the update.

I didn't even realize this could be a problem despite the next paragraph implying it's very well known.

Some thing I've learned, but rarely seen explained anywhere: Storing the vectors is most likely not an issue, mostly likely you're having a problem with the indexes on top of them in which case you can use quantized vector indexes[0] (handled by pgvector) which will get past the limits imposed by PostgreSQL.

I had to switch off pgvecto.rs at some point and figured that out.

I don't have specific experience with the Azure environment here, but this probably applies if you have access to pgvector.

[0]: Types of indexes + number of bits supported at bottom of this section: https://github.com/pgvector/pgvector#hnsw

In my experience the qwen models are best locally, but gemma ones have always been good. gemma4 is a notable improvement.

Oh I'd never connected this. It makes so much sense. I'd always wondered what Texas had to do with computing that made so many things start there.

It's a shame that it is such a niche show in practice. The acting of Lee Pace and Mackenzie Davis in particular are so good across all 4 seasons.

I recommend it at every chance I get, but few people ever watch it. They're more likely to give Silicon Valley a try.

I used the nftables Wiki to learn all the basics I know about nft: https://wiki.nftables.org/wiki-nftables/index.php/Main_Page

Here is their example relevant to the current article: https://wiki.nftables.org/wiki-nftables/index.php/Simple_rul...

My main reason is that nft applies configs atomically. It also has very good tracing/debugging features for figuring out how and why things aren't working as expected.

That said, I think many distros are shipping `iptables` as the wrapper/compatibility layer over nft now anyways.

Agreed. NixOS + Tailscale is 99% there for me. Using Claude Code to deal with whatever other package I need built with nix while I'm working on $day_job things helps get me to a fully working system. Besides the fact that running containers via podman or docker (your choice) is super easy via a NixOS config.

Combine that with deploy-rs or similar and you have a very very stable way to deploy software with solid rollback support and easy to debug config issues (it's just files in the ./result symlink!)

I challenge anyone to try building a C compiler without a big suite of tests. Zig is the most recent attempt and they had an extensive test suite. I don't see how that is disqualifying.

If you're testing a model I think it's reasonable that "clean room" have an exception for the model itself. They kept it offline and gave it a sandbox to avoid letting it find the answers for itself.

Yes the compression and storage happened during the training. Before it still didn't work; now it does much better.

I didn't personally experience it (I was too young), but I think that was part of "the mission" since pre-9/11. The point of the ID check is to make sure the boarding ticket and ID match.

In effect that tracks who is going where.

They have already renamed again to openclaw! Incredible how fast this project is moving.

I can't open the exterior doors of my apartment without an app (despite having an option for RFID dongles. they heavily advise against it)

At least I need my apple watch with cellular enabled so I can dial myself in.

Maybe but it's a common platform and explains why IPv6 support still isn't fully there.

Ah yes, and convincing friends/family/partners to use Signal instead of Whatsapp clearly what will convince them is that they need to setup, acquire, and use cryptocurrency to register or connect with me on the encrypted messaging service. "No thanks, I just use Whatsapp/iMessage. I heard they're actually e2e encrypted too, so what's the problem?"

Management then cared that their one chance would work. Today management just wants it to mostly work.

Incentives and goals are very different between the two. We could very much build even more incredible things today; and would argue that we actually do. Just only in the places that seem to matter enough to do that type of special effort for.

I get it, but part of the reason they're on motorcycles is cars are too expensive for them. Would making it safer and costing $20-30k base like it does in America be more or less ethical?

I don't think there is a clear answer. Safer car means owners of the car survive more and more often: which is unequivocally great. Making it cheaper means there are more people in a safer vehicle than before which can reduce their likelihood of getting hurt/dying, but they're not as protected as they could be. This is, on the whole, also really good.

It just so happens car makers pick to make it cheaper. It's good for business and good for those who couldn't afford it before. It is bad for those who encounter the situations where additional protection would have saved them.

Unfortunately it simply is true. The "same car" in another country is made cheaper. There are a variety of ways to do it and often it's justified as better than the alternative. For instance in India the goal is getting people off of motorcycles as that is a huge cause of driving deaths. To do that they remove various airbag systems, auto braking systems, and etc. which are not required in India... but are required in America or the EU.

Even between the EU and America there are differences in regulation with the EU often getting the stronger regulation first.

This video is my source for all of this: https://www.youtube.com/watch?v=dVI-vFq39-I

I've done my best to remember most of it.

No one was, nor was implied to be, deceptive. They described it in their own words and the word they chose is clearly inaccurate or misused.

That's all.

They could have described it as "game-like design" or something else.

Ads pay for everything and Google/Meta are making huge profits (minus AI spending)... so probably most people.

I'm sure models which replace characters in realtime will also become popular. I would imagine some company thinking it would be cool if the main character looked slightly more like whatever main audience it's being shown to and it's done on their playback devices (so, of course, it can be customized or turned off).

I find the idea fun, kinda like using snapchat filters on characters, but in practice I'm sure it'll be used to cut corners and prevent the actual creative vision from being shown which saddens me.

Yes, I don't see how else they'd make any difference or get close to their goal without waiting 10 years for the computers to be slowly replaced.

Look at the latest changes in iPhone hardware for security. Some are saying the iphone is unhackable now.

There are no reasons such improvements couldn't work for Windows and I am guessing this is their best idea for raising the floor of hardware protection they can assume is available.

I agree the push, which is quite disruptive, must be influenced by someone higher up. But I think its as simple as "Windows is insecure" is something they want to make untrue.

I think the article's point is that many on Windows 10 would upgrade to 11 if they could. Their computers are simply not supported and to get around that requires some technical skills.

Feels like the next logical move to me: they need to build and grow the demand for their product and API.

What better than companies whose central purpose is putting their API to use creatively? Rather than just waiting and hoping every F500 can implement AI improvements that aren't cut during budget crunches.

Sam Altman said in an on-the-record dinner interview with Platformer[0] that besides R&D ChatGPT was breakeven and Brad Lightstep, head of ChatGPT, corrected him by saying they were close, but not yet break even.

I assume Sam and Brad both understand the unit economics of their product.

Article is pay-walled for me, but I heard it on their podcast[1]. Which somehow I heard fine, but that page is getting pay-walled for me.

[0]: https://www.platformer.news/sam-altman-gpt-5-interview-light... [1]: https://www.nytimes.com/2025/08/15/podcasts/hardfork-gpt5-pe...

Just because it's not independent and autonomous does not mean it could not be intelligent.

If existing humans minds could be stopped/started without damage, copied perfectly, and had their memory state modified at-will would that make us not intelligent?