Seems to me like it was an explicit design decision that fixed some use cases and broke others, rather than a regression in the sense of "A bug we introduced but didn't catch." As such a regression test wouldn't have caught this.
Someone made a design decision that broke some use cases.
The last token ring standard appeared in 2001 and there has been a huge amount of network-technology development since then. Tokenring is tiny and incredibly niche. It may as well be dead.
Ethernet got its latest in 2022 and another is in the pipeline.
Mike Levine (a financial blogger for Bloomberg) says the industry is calling these sorts of transactions "reverse acqui-hires" to distinguish them from a normal (forward) acquihire where the acquirer buys the whole company even if all it wants is the people. He discusses some of the advantages and legalities here:
This reflects pretty much every result of mine from Google these days. Not just navigation, but search too.
The only thing I can figure is that Google's consumer-facing free-to-use models are set to "reason as fast as possible with only the most basic plausibility checks." I have to correct them like this every single time.
On Android, the public APIs available for digital assistants are more limited than the ones that Google gets to use for their own products. So if you want a digital assistant and you'd like to be able to hold a button on your phone and have it open a thread with an LLM (even if your screen is off), your digital assistant has to be Gemini. You can configure ChatGPT or Claude in its place, but then they can't manipulate or interact with apps in your phone very much. Or you can use an app like that for all of your LLM use but you can't conveniently summon it.
Similarly, you can run local LLMs on your Google Pixel smartphone, but if you want them to be hardware accelerated using the chip that's included for that very purpose on your phone... it has to be Google's models via Google's apps. So you can't say "alright, a fast, local model which sounds great, but I want one that is trained to always run a search before answering, in a harness with good search tools for answering the kinds of practical questions I often ask", you're actually not free to put that together in a practicable way.
So there is indeed some competition here. But it's not really fair competition.
It's a little bit like any benchmark, especially microbenchmarks. It is useful as far as it reflects the broader use cases, but the risk of overfitting to the specific task and not the broader ones is very high.
> "Everybody just complains ... and ... we say it's taking energy away from GIMP."
It's almost like Hacker News is a giant group of people, and different subsets of people comment on different subsets of articles (aka, the Goomba Fallacy).
I would like to think that too, but the illusion was shattered in the past few years with AI comment spam and reddit-style injection of politics and astroturfing.
A lot of software developers are seduced by the old “80/20” rule. It seems to make a lot of sense: 80% of the people use 20% of the features. So you convince yourself that you only need to implement 20% of the features, and you can still sell 80% as many copies.
Unfortunately, it’s never the same 20%. Everybody uses a different set of features.
Someone made a design decision that broke some use cases.
reply