In that case it like someone controlling the DNS records for a banned site could cause some mischief

How is a company like mintlify getting so many big name customers for what appears to be a static site generator + hosting? Is there some secret sauce I'm missing, what is the value proposition?

fun fact: last BigCo I worked in had an elaborate architecture/security bar for new applications/features but offered a clever workaround - you could use a pre-approved solution and skip numerous quality checks and approvals, so every single PO was pushing for that specific solution.

The result? A static html with 500 ppl audience was billing a whooping 2k EUR a month, because that was the cost of that pre-approved architecture.

Best part - I was championing a company wide solution for that problem for over a year, which resulted in board level special operation with 100k budget only to get that budget snugged by people couple steps above the ladder.

Lots of these companies are YC companies, and they tend to use other YC products. For those that aren't, its easier to just use what other big names are using, and having YC as a backing name is quite useful in that regard.

Convenience and developer uncertainty. I fall pray to the "it's paid, so it must be better" fallacy, and the "they know what they are doing, they are pros" illogicality.

I genuinely don't know, especially for Vercel to be using them. Vercel themselves can easily be used to host static-ish documentation.

But it looks like Mintlify are using Vercel on the backend: https://vercel.com/blog/mintlify-scaling-a-powerful-document...

So it's just a Vercel wrapper?

It's not clear if you need it on both ends to get an advantage?

The bottleneck in SSH is entirely on the receiving side. So as long at the receiver is using HPN-SSH you will see some performance improvements if the BDP of the path exceeds 2MB. Note: because of changes made to OpenSSH in 8.8 the maximum buffer with OpenSSH as the sender is 16MB. In an HPN to HPN connection that maximum receive buffer is 128MB.

There's a whole industry of people selling solutions to WordPress's failings, all of whom have strong incentives for it not be properly improved.

A bit like psDooM

I always assumed high end CCTV cameras already did something like this?

web search too is off by default

H100's can be $2 and hour, so $192 an hour for the full cluster. They report 22k tokens per second, so ~ 80 million an hour, thats $16 an hour at $0.2 per million. Maybe a bit more for input tokens, but it seems a long way off.

I think you mis-read. Thats 22k tokens per second per node, so per 8 h100's. With 12 nodes they get 264k tokens per second, or 950 million an hour. This get's you to roughly $0.2021 per million at $2 an hour.

What is obnoxious is that certificate transparency logs mean that you now have to effectively centrally register any new domain you put online. That means you instantly see a whole load of traffic to your domain from bots, scrapers, beg bounty scanners etc. Any new site has to be designed to handle that baseline of traffic.

I understand the point of CTL's and it's necessary given that every browser and device is configured to trust CA's that you wouldn't actually trust. It's had awful side effects for people who want to host low traffic sites, or fly under the radar for whatever reason.

Point taken.

One option to avoid this could be to use the DNS-01 challenge to get a wildcard cert from Let’s Encrypt. Then CT will not expose your subdomains.

Two problem with offering usage limits is the real limit you could offer if all users hit it, is low. The users with usage far below the limits feel they are getting a bad deal, compared to if they can't see the limits and they don't hit them, they feel they have "unlimited!".