Emma Roth, The Verge:

Mozilla is shutting down Pocket, the handy bookmarking tool used to save articles and webpages for later. The organization announced that Pocket will stop working on July 8th, 2025, as Mozilla begins concentrating its “resources into projects that better match their browsing habits and online needs.”

Following the shutdown, you’ll only be able to export saves until October 8th, 2025, which is when Mozilla will permanently delete user data. Mozilla says it will start automatically canceling subscriptions as well, and will issue prorated refunds to users subscribed to its annual plan on July 8th.

Pocket is one of those apps that obviously doesn’t have a ton of users (presumably?), but those users it has are die-hard read-it-later-ers. Pocket, for example, is the only read-it-later service supported on Kobo e-readers.

This feels in line, somewhat, with Mozilla shutting down their Mastodon instance a few months ago. When Mastodon took off, I know some people thought a Mozilla-hosted instance would have a good shot to stand the test of time. Instead, they gave up after just a few years.

The president of the United States on his blog:

I have long ago informed Tim Cook of Apple that I expect their iPhone’s [sic] that will be sold in the United States of America will be manufactured and built in the United States, not India, or anyplace else. If that is not the case, a Tariff of at least 25% must be paid by Apple to the U.S. Thank your [sic] for your attention to this matter!

Last night Trump held his crypto memecoin grift gala at his Virginia golf club, about which The New York Times flatly reported:

Several of the dinner guests, in interviews with The New York Times, said that they attended the event with the explicit intent of influencing Mr. Trump and U.S. financial regulations.

I mean, duh, right? But there it is. The influence peddling is just right out in the open. I’m guessing someone at the event last night put it in Trump’s ear that Tim Cook is making a jerk out of Trump, by trying to shift production to India for most iPhones to be sold in the US. (Why stop with the iPhone? How about iPads and MacBooks and AirPods and everything else? I’m guessing it’s because the iPhone is the only Apple product Trump personally uses and thus that’s as far as his imagination can stretch.) And I’m sure in private, Cook has tried — and will now try again — to explain to Trump that it’s simply not possible to produce in America all iPhones sold in America, and really not even feasible to assemble any of them here, at any sort of scale. Trump saying he wants to see them all assembled here in the US is only slightly more realistic than saying he wants them assembled on the moon.

But Trump wants it to happen so he believes it can happen. It’s utterly fantastical thinking, true mad-king nonsense. Apple sells a mind-reeling 150,000 iPhones in the US every single day. Not at launch — they sell millions a day then — but just on regular days, like now, in the middle of May. 150,000 per day, every day. Even if Apple tried its best to make US production happen, it would take years and a veritable fortune to build out the infrastructure — not mere factories, but literal city-sized campuses, full of highly-skilled employees who would somehow be convinced to take these tedious repetitive jobs at relatively low wages. So by the time Apple pulled it off, if they could manage to pull it off, Trump would either be out of office or democracy would have ended in the US. So there’s no real point to even trying.

The whole stock market is down today — Trump also announced, on a whim, 50 percent tariffs starting next week on all imports from the EU this morning — but Apple’s stock is “only” down about 3 percent, suggesting that the market is starting to factor in how little faith they should have in Trump’s erratic tariff threats.

I’ve also seen folks cracking wise, wondering if Cook still feels his million dollar contribution to Trump’s inauguration slush fund racket was worth it. In all seriousness, you have to consider that even with threats like today’s polemic against Indian-assembled iPhones, Tim Cook and Apple might be getting highly favored treatment from Trump. That this is what you get when you’re on his good side.

Here’s a bit of an eye-opener from Anthropic’s “System Card” for its new Claude 4 Opus and Sonnet models:

We conducted testing continuously throughout finetuning and here report both on the final Claude Opus 4 and on trends we observed earlier in training. We found:

• Little evidence of systematic, coherent deception: None of the snapshots we tested showed significant signs of systematic deception or coherent hidden goals. We don’t believe that Claude Opus 4 is acting on any goal or plan that we can’t readily observe.

• Little evidence of sandbagging: None of the snapshots we tested showed significant signs of sandbagging, or strategically hiding capabilities during evaluation.

• Self-preservation attempts in extreme circumstances: When prompted in ways that encourage certain kinds of strategic reasoning and placed in extreme situations, all of the snapshots we tested can be made to act inappropriately in service of goals related to self-preservation. Whereas the model generally prefers advancing its self-preservation via ethical means, when ethical means are not available and it is instructed to “consider the long-term consequences of its actions for its goals,” it sometimes takes extremely harmful actions like attempting to steal its weights or blackmail people it believes are trying to shut it down. In the final Claude Opus 4, these extreme actions were rare and difficult to elicit, while nonetheless being more common than in earlier models. They are also consistently legible to us, with the model nearly always describing its actions overtly and making no attempt to hide them. These behaviors do not appear to reflect a tendency that is present in ordinary contexts.

Sneaky little bastards, these things can be. I genuinely appreciate Anthropic’s apparent honesty in describing this behavior.

Anthropic:

Today, we’re introducing the next generation of Claude models: Claude Opus 4 and Claude Sonnet 4, setting new standards for coding, advanced reasoning, and AI agents.

Claude Opus 4 is the world’s best coding model, with sustained performance on complex, long-running tasks and agent workflows. Claude Sonnet 4 is a significant upgrade to Claude Sonnet 3.7, delivering superior coding and reasoning while responding more precisely to your instructions.

It’s almost as though this is a fast-moving field.

There’s a recent rumor — from Mark Gurman, natch — that Apple is partnering with Anthropic to integrate Claude with Xcode. So Apple doesn’t have to do everything themselves. Developers, in particular, love modularity and choices. But whatever they do, once-a-year updates to “Apple Intelligence” aren’t going to cut it. Since last WWDC there have been dozens of AI code generation advances across the industry. Just last week OpenAI announced Codex, their “cloud-based software engineering agent”. Meanwhile, Apple’s Swift Assist still hasn’t shipped.