pajamasam

This. If they had the data access logs enabled, they would've been able to confirm whether or not the data was accessed.

Maybe a stolen wallet?

So much for independent journalism.

I've been wanting to try it. In what way is it finnicky?

It's also all open source and local-first. You can also add your own self-hosted WebDAV server for syncing across clients. Also, it's developed by a nonprofit organization.

If there are more doctors, it would require less of each one's time though.

Don't know when he bought his devices, but Apple Intelligence is not yet 2 years old so it might no even have existed when he bought his Mac.

He did justify it in https://dbushell.com/notes/2026-09-16T05:01Z/

I think your unpilled and lower tier pills are often disguised anti-pills. Anti-pilled as in AI is morally bad. I count myself as both ASI pilled and anti-pilled.

I'd be interested in what data they're using in the first place. It looks like their data is behind a paywall.

Also, funny that their page with for "Companies running Cloudflare" says there are only "887 European organisations using Cloudflare."

None of the most read books of 2026 so far according to Goodreads are significantly older (https://www.goodreads.com/book/most_read?category=all&countr...). Also not the most sold Amazon books: https://www.amazon.com/gp/bestsellers/2026/books

It’s how they make money

longreads.com, my RSS feeds in NetNewsWire, and podcasts mostly

Apparently we can because a lot (most?) of us are still using GitHub even after all their outages recently.

> anyone saying that’s a sign of LLM use should not be listened to

You don't need to believe me. You can read the studies about its statistically higher occurrence in LLM writing vs human writing (https://arxiv.org/pdf/2603.27006) or you can do the analysis yourself.

Also, no one said you should stop using them. But overusing them, along with emulating some of the other common traits of LLM generated writing, will give people the impression that you didn't bother to write something yourself.

It contains four emdashes, it overuses the Rule of Three (https://en.wikipedia.org/wiki/Wikipedia:Signs_of_AI_writing#...), it has the typical "not X, but Y" sentence, it unduly emphasizes the significance of Buqtrack (https://en.wikipedia.org/wiki/Wikipedia:Signs_of_AI_writing#...) ("preservation for the ages", really?).

Also, in my experience, LLMs seem to love to say something went "dark" or "silent", to mention a "generation" of people, and to say something "matters". "no corporate filter" also seems like a strange thing to say.

Kagi does something like this: https://blog.kagi.com/slopstop

LinkedIn also added a button to report AI slop, but I'm doubtful it would be of much help on LinkedIn since every second post seems to be written by an LLM.

From the commits, it looks like it only took OP about 4 hours with Cursor.

Hmmm no, Meta is making money from businesses using their API.

And a single person's answer to a statistical phenomenon is going to help you how?

Interesting idea! A demo video would be great :)

1. They still make the data freely available. 2. Hosting the data is not free.

I honestly thought the whole thing was satire and that that line was a riff on OpenAI.

You think other CDNs are immune from the kinds of issues Cloudfare has?

> I wonder can you just set up fallback to origin in DNS

Then you origin needs to be publicly accessible and you lose the DDOS protection that a CDN gives you.

> All the other ones.

They're just as susceptible to outages as Cloudflare is.

Anthropic says they may not train their models using your data, but apparently Persona (the service they will use for identity verification) WILL according to https://thelocalstack.eu/posts/linkedin-identity-verificatio...

Persona also might send your data to 17 different subprocessors (16 if you exclude Anthropic itself).

There are multiple good reasons to use a CDN though. What’s your suggestion for an alternative?

The article mentions some of the places use Yondr phone pouches.

Oh, I'm sure you have, mr bot