loudwhisper

Kagi is an engine, searx is a meta-engine. That's what I meant. Which means kagi does not simply collate results from multiple source (like searx does), but implements its own logic. This means that - for example - it deranks website with many trackers, or can implement various features on top of the results. So it's not a nitpick, it's a substantial difference between an engine (kagi) and a metaengine (searx), which is essentially a proxy + aggregation of other engines.

It's a known fact that brave optimizes result based on google data, and the kagi guys themselves in fact added that - with it being cheaper than google API - it could be a vector to eventually reduce cost for google API without impacting results.

That said, AFAIK kagi does not pose as a nonprofit, I think they make extremely clear that running searches (scraping, paying API, etc.) cost money and that they need to be profitable. Their stance is that by using a subscription model, their business interests align with user's interests of providing good searches, rather than results that benefit advertisers, which is completely reasonable. This is literally written in their "why pay for searches" article that is presented when they show the pricing.

Of course it is a big difference, and you can argue for pros and cons of both options. I personally think the internet should not be based either on megacorp nor on free labor. Would I prefer kagi being a co-op? Sure. But it's not like relying solely on free labor is free from any moral implication either (sure, you can donate, and I do to Lemmy for example, but only a minority does).

That's not an engine, it's a metaengine. The results are still tied to the engines used, which means if they are trash, you get trash. Kagi uses a mix of google/yandex/brave etc. and then elaborates them as well, in addition to have their own scraper for things like the small web (which is great to surface personal blogs).

They are not comparable. Also, kagi's privacy policy is exemplar and the account can be paid in crypto now (if you don't want to use CC).

Besides, there is no such thing as free hosting, similarly to Lemmy, it's just someone paying.

Did you mean western?

They did not disclosing any content of any email. They disclosed the very little they have. Once they have been forced to log IP addresses and that was turned to law enforcement, another time they were forced to disclose a recovery email address. These facts if anything should help build trust in proton, as they show how little they collect and therefore can disclose. With signal is the same, they collect super minimal info (the time you last logged in and a couple more data points, I think), and that's what they disclosed in the past.

It's a non-news.

The law - for good or for bad - is what defines rights. If there is a judge which says that an investigation has to happen, and also the companies ensured that the claim is legit (you see from the stats that the context 15-20% of the data requests), then what else can be done?

You cannot operate illegally, so either you comply or you shut down.

I run Prometheus on a separate cluster, so I plug my servers with node_exporter and scrape metrics. I then alert with grafana. To be honest, the setup is heavier (resource usage-wise) than I would like for my use case, but it's what I am used to, and scales well to multiple machines.

I have seen this post and decided to respond via a separate blog post. https://loudwhisper.me/blog/containers-isolation/

The short answer is that yes, they do. And yes lowering the privileges of the user helps in avoiding container escapes, which basically makes the other advantages for containers valid. You can, however, achieve the same using (relatively obscure, imho) systemd settings, running with flatpak etc. Namespaces + Cgroups + Seccomp + Capabilities = better security. Containers make it easy to use all of the above.