Right, that's the lesson being learned here today.

To be fair, up 'till now, afaik these services actually have been trustworthy on this count, so this is a new low.

Just because people extend you a lot of trust to do A,B, and C doesn't mean you can't lose it if you randomly start doing D instead.

I don't think we can run a viable (high trust) western economy if we constantly have to beware of our own providers. Principle of least surprise applies.

One day the delivery service starts adding free sausages to every delivery on an opt-out basis.

By the time you realize, they'll already have delivered adulterated meals. You won't have a lot of customers left!

People are often vegetarian for ethical reasons. In a similar way, many people have Ethical Opinions (tm) about Analytics as well. They don't like it when other people collect private information about themselves, and refuse to Do Unto Others.

But even if you're fine with analytics in general: When intermediaries and other transitive dependencies (like software libraries) start injecting analytics on an opt-out basis, you simply can't control what's going on. And that's before we talk about the ethics and legality of third parties modifying content in-flight.

In the EU, the law actually codifies the ethics at play here (GDPR). Several people point out that cloudflare disables this 'feature' for the EU. This makes a lot of sense, because it would have been a huge mess for everyone impacted to have to go to their customers with a "sorry, we leaked your PII, here's how we'll make you whole"; as well as possibly lawsuits against cloudflare to recover those damages upstream.

https://en.wikipedia.org/wiki/Natural_law

I had something similar happen, but with a different outcome. One week I noticed I had some mild distortions in one eye I which I thought were a bit funny, one of those things, you know?

No pain or anything, but eh, the next sunday morning I was just messing around, and just for the heck of it, I described the funny distortion thing to Claude Opus... which proceeded to flip out and told me to call a doctor Right Now.

Clearly Claude was broken, right? And what a weird failure mode! It just kept circling back to telling me to see a doctor "please just call at least!".

I deleted the conversation (something I now regret) and resolved to get on with my day.

But... you know, small doubt having surfaced, I decided to use the expert system provided by my local doctor's practice;

which also told me to call immediately

... and that's how I ended that Sunday morning

    in the passenger seat of a Porsche 

    racing down the autobahn to a nearby hospital.

Somehow this sentence is self-contradictory, but I can't quite lay my finger on the exact spot.

I surmise that most people don't show you their slop/practice work, else the world would already be drowning as is. The few that actually do put their bad early work out there dominate the landscape though.

It turns out that some people get past the practice stage. Maybe most, eventually, hopefully?

Meanwhile, let's leave the hobby-horses in the closet for now. I won't comment on people's programming tool preferences.

One important reason is due to Permission Fatigue: Of course you check everything! You're diligent! The last 100 requests were all ok, so you're down to hitting yes, yes, yes, yes, yes, yes, yes, yes ...

... oops, that third yes should have been a no!

Are you objecting to terminology here?

Are you proposing we say "Fable-In-Claude-Code tried..." instead?

Hmmm... something like that might be necessary. Sure we should typically be tolerant of loose language; but people do keep referring to wildly different contexts in ai conversations, and end up talking past each other.

Running gemini on web is a genuinely different experience to running Fable in claude code, different again from GPT-5.6 in openclaw, or in an ide or etc ...

This could very easily be turned into a preservation and archiving operation with just a tweak of the laws, or a carve-out.

And make the bank once, and make it legal to train on? How people in the bank get compensated is a different question, but -while almost impossible to settle on an individual basis- could be settled reasonably in bulk by some form of mandatory implied statutory contract?

You wouldn't trust LLM generated software without unit tests either.

(Maybe not even with just the unit tests, but you've got to start somewhere.)

Either way, pointing the finger at your tools is a cheap move. Don't blame the chainsaw for your own negligence!

As someone originally coming off of RISC OS[1] way back when, the mac way is closer to intuitive than the windows way, while Linux's approach of splitting files across the entire FS so that you can't possibly find all of them is "The Real WTF".

(Nixos is even further from that original intuition in a lot of ways. Oh for the days of drag-dropping the !App dir to wherever you happen to have space; and maybe dragging it to !menon if you needed to actually find it back )

[1] The operating system that came with a thin welcome booklet with reassuring pencil art explaining how to hold a mouse, and what the difference between a click and a doubleclick was. [2]

[2] https://archive.org/details/risc-os-5.28-user-guide/page/6/m... Oh wow, still exists in the RISC OS 5 manual. This kind of manual used to be famed : Eg. Chapter 1 introduces what a desktop is; Chapter 2 explains the mouse; 3 Explains the keyboard ; 4 explains what a window is. Give a motivated naive user this guide and there's a small chance they might actually grok the thing when they're done.