We should treat LLM somewhat like patents or drugs. After 5 years or so, the models should become open source. Or at very least the weights. To compensate for the distilling of human knowledge.

people point to the rare bug report that deletes absolutely everything in the account. but at this point, I don't even know if it's true.

I've been hit by that bug, although it only deletes mail AFAIK. There's a separate bug that completely corrupts the mail database on compaction, making Thunderbird lock up including for every future launch.

Its a beautiful open source effort but products that have bugs like that languish for 10-20 years just aren't reliable. I need my mail client to be reliable.

I've been using it to close to 20 years with multiple accounts and it was rock-solid. I wouldn't extrapolate from anecdata, in either direction.

But we should not spread FUD. If you can link to the bug I'd be interested, otherwise it doesn't add much value to claim this.

Yes, FUD and long held myths can be found anywhere. But speaking as a staff member and someone who has seen first hand user reports, here is some straight shooting:

* there are rare cases of a profile either misplaced (exists but not correctly pointed to) or gone - it is something which I understand Firefox people are working on (Thunderbird uses the Firefox profile system) * there are extremely rare reports where prefs.js is corrupted * there are no compact failures in current versions - there are no open bug reports for recent versions, so it has been totally obliterated by a rewrite and subsequent fixes. Most user reports of compact failure are attributed to other causes of folder corruption * folder corruption can occur as easily from external sources as from product bugs.

Anyone who has a problem can file a support request at https://support.mozilla.org/en-US/questions/new/thunderbird to get assistance.

Also, beware drawing broad conclusions about other users' experience from one's own personal experience. I have almost never experienced corruption - once in the last 10 years. But I am also using a Thunderbird profile that has gone through 5 different laptops, two different OS, using daily builds, which is AMPLE opportunity to have had multiple catastrophic failures. But because I know other users experiences I consider myself lucky.

There are lots of bugs and the person you're responding to might have a different one in mind, but this is the classic, ancient unresolved bug:

https://bugzilla.mozilla.org/show_bug.cgi?id=462156

Although in recent years it looks like it turned from a bug about one specific (never resolved) issue to a more general troubleshooting session related to data loss issues.

I used to maintain a mailbox in dropbox that tended to work across my mac, linux and windows environments... it was pretty great... at some point a few incompatible releases across the environments broke everything and had other bugs that I could no longer revert from. I pretty much haven't touched it in a while.

Anthropic bonuses seem to be code for "your usage limits are going down soon."

Literally every time

Are you using Pi with a cloud subscription, or are you using the API?

I can't imagine the logic chain that made you come to that conclusion.

There's a big difference between non-compliant, illegal, and criminal.

But increases credibility?

Sure, but the subtasks don't feel completely disparate, probably because there's shared context in working memory.

$100 is for casual usage. To use it for work requires $200/mi plus extra usage.

CC does all of my work and I have hit the $100/mo Max plan limit only once, and I probably could have just called it quits for the day at that point.

Are you saying that is the Claude rules for the subscription or saying because work uses more of it?

I ofter wonder if I'm missing something, but shouldn't we be able to edit the context manually???

In that way we could erase prompts and responses that didn't yield anything useful or derailed the model.

Why can't we do that?