Linked List: August 2025

Wednesday, 6 August 2025

OpenAI:

We’re releasing gpt-oss-120b and gpt-oss-20b — two state-of-the-art open-weight language models that deliver strong real-world performance at low cost. Available under the flexible Apache 2.0 license, these models outperform similarly sized open models on reasoning tasks, demonstrate strong tool use capabilities, and are optimized for efficient deployment on consumer hardware. They were trained using a mix of reinforcement learning and techniques informed by OpenAI’s most advanced internal models, including o3 and other frontier systems.

The gpt-oss-120b model achieves near-parity with OpenAI o4-mini on core reasoning benchmarks, while running efficiently on a single 80 GB GPU. The gpt-oss-20b model delivers similar results to OpenAI o3‑mini on common benchmarks and can run on edge devices with just 16 GB of memory, making it ideal for on-device use cases, local inference, or rapid iteration without costly infrastructure. Both models also perform strongly on tool use, few-shot function calling, CoT reasoning (as seen in results on the Tau-Bench agentic evaluation suite) and HealthBench (even outperforming proprietary models like OpenAI o1 and GPT‑4o).

Simon Willison:

The long promised OpenAI open weight models are here, and they are very impressive. [...]

o4-mini and o3-mini are really good proprietary models — I was not expecting the open weights releases to be anywhere near that class, especially given their small sizes. That gpt-oss-20b model should run quite comfortably on a Mac laptop with 32GB of RAM.

Tuesday, 5 August 2025

Anthropic Releases Claude Opus 4.1 ★

Anthropic:

GitHub notes that Claude Opus 4.1 improves across most capabilities relative to Opus 4, with particularly notable performance gains in multi-file code refactoring. Rakuten Group finds that Opus 4.1 excels at pinpointing exact corrections within large codebases without making unnecessary adjustments or introducing bugs, with their team preferring this precision for everyday debugging tasks. Windsurf reports Opus 4.1 delivers a one standard deviation improvement over Opus 4 on their junior developer benchmark, showing roughly the same performance leap as the jump from Sonnet 3.7 to Sonnet 4.

Nothing spectacular here, but incremental improvements add up. Mike Krieger — best known as a co-founder of Instagram, now chief product officer at Anthropic — in an interview with Bloomberg:

“In the past, we were too focused on only shipping the really big upgrades,” said Anthropic Chief Product Officer Mike Krieger. “It’s better at coding, better at reasoning, better at agentic tasks. We’re just making it better for people.” [...]

“One thing I’ve learned, especially in AI as it’s moving quickly, is that we can focus on what we have — and what other folks are going to do is ultimately up to them,” Krieger said when asked about OpenAI’s upcoming release. “We’ll see what ends up happening on the OpenAI side, but for us, we really just focused on what can we deliver for the customers we have.”

I’m on board with the idea that Apple need not acquire any of these AI startups, but if they do, Anthropic — not Perplexity — seems the one most aligned with Apple’s values. And I don’t mean values in just an ethical sense, but their entire approach to product development in general.

Google Dunks on Apple Intelligence in New Pixel 10 Ad ★

Tom Warren:

In a new Pixel 10 ad, Google dunks on Apple’s failed promise of Siri AI improvements, with a narrator that suggests you could “just change your phone” if you bought “a new phone because of a feature that’s coming soon, but it’s been coming soon for a full year.”

The 30-second spot appeared on YouTube and X today, teasing the launch of Google’s new Pixel 10 devices on August 20th.

The whole Siri/Apple Intelligence thing has been an enormous self-inflicted embarrassment, but when it comes to Pixel phones, all I can think of is that Mad Men “I don’t think about you at all” GIF.

Lawsuit Alleges That Meta Pirated and Seeded Massive Amounts of Porno for Years to Train AI ★

Ashley Belanger, writing for Ars Technica:

Porn sites may have blown up Meta’s key defense in a copyright fight with book authors who earlier this year said that Meta torrented “at least 81.7 terabytes of data across multiple shadow libraries” to train its AI models. [...]

After authors revealed Meta’s torrenting, Strike 3 Holdings checked its proprietary BitTorrent-tracking tools designed to detect infringement of its videos and alleged that the company found evidence that Meta has been torrenting and seeding its copyrighted content for years — since at least 2018. Some of the IP addresses were clearly registered to Meta, while others appeared to be “hidden,” and at least one was linked to a Meta employee, the filing said.

According to Strike 3 Holdings, Meta “willfully and intentionally” infringed “at least 2,396 movies” as part of a strategy to download terabytes of data as fast as possible by seeding popular high-quality porn. Supposedly, Meta continued seeding the content “sometimes for days, weeks, or even months” after downloading them, and these movies may also have been secretly used to train Meta’s AI models, Strike 3 Holdings alleged.

The porn site operator explained to the court that BitTorrent’s protocol establishes a “tit-for-tat” mechanism that “rewards users who distribute the most desired content.” It alleged that Meta took advantage of this system by “often” pirating adult videos that are “often within the most infringed files on BitTorrent websites” on “the very same day the motion pictures are released.”

Meta is an empty husk of a company with no values, no beliefs, other than growth and dominance for the sake of growth and dominance.

Ghost 6.0 ★

Ghost:

When we announced Ghost 5.0 a few years ago, we were proud to share that Ghost’s revenue had hit $4M — while publisher earnings had surpassed $10M. It felt great to have such a clear sign that our goal to create a sustainable business model for independent creators was succeeding.

Today, Ghost’s annual revenue is over $8.5M while total publisher earnings on Ghost have now surpassed $100M. [...]

Unlike our venture-backed peers obsessed with growth at all costs, we’re structured as a non-profit foundation that serves publishers directly with open source software. We believe independent media cannot be beholden to proprietary tech companies, so Ghost publishers don’t just “own their email list” — they own the entire software stack that underpins their business, end to end.

Not a centralized platform controlled by a single corporation, but open infrastructure that’s shared by everyone.

Aside from my feelings about Substack — clearly the main target of Ghost’s shade-throwing here — it’s just great to see so many indie publishers and writers thriving on Ghost.

Regarding Those Rumors of Apple Pursuing an Acquisition of Perplexity ★

MacRumors, on June 20:

Apple executives have been discussing the possibility of the company making a bid to acquire Perplexity AI, according to Bloomberg’s Mark Gurman. Perplexity is one of the leading AI startups that has proven popular as an AI-infused web search engine.

From that Bloomberg report by Gurman:

Adrian Perica, the company’s head of mergers and acquisitions, has weighed the idea with services chief Eddy Cue and top AI decision-makers, according to people with knowledge of the matter. The discussions are at an early stage and may not lead to an offer, said the people, who asked not to be identified because the matter is private.

Such a deal would help Apple develop an AI-based search engine, part of efforts to cope with the potential loss of a longstanding arrangement with Google. That partnership, which involves making Google the default browser on devices, generates roughly $20 billion a year for Apple — and is now under threat from US antitrust enforcers.

To date, Apple executives haven’t discussed a bid with Perplexity management. Bloomberg News reported earlier Friday that Meta Platforms Inc. tried to buy Perplexity earlier this year.

“We have no knowledge of any current or future M&A discussions involving Perplexity,” the AI startup said in a statement. Apple declined to comment.

I think, reading between the lines of Apple’s prepared remarks and Tim Cook’s and CFO Kevan Parekh’s answers to analyst questions last week after announcing quarterly earnings, that it doesn’t sound like they believe Apple needs to make a big acquisition in this space. Apple could probably acquire Perplexity for a lot less than it would cost to acquire other companies in the space, but that’s partly because Perplexity doesn’t develop or train its own models. Perplexity primarily puts up its own front end atop models from OpenAI, Anthropic, Mistral, Google, and xAI. I really don’t see what buying Perplexity would gain Apple.

But even putting that aside, it just seems like Perplexity is sketchy. This whole thing where Cloudflare seemingly caught them redhanded ignoring robots.txt directives and masquerading their user-agent makes the company seem like a poor cultural fit for Apple. I can see why Meta, a company without a moral compass, approached Perplexity to sniff around regarding an acquisition. That seems like a good cultural fit.

I can’t see why Apple would want to get involved with a company like this though. Gurman’s report makes it sound like his sources are inside Apple, but man, this “Apple + Perplexity” thing feels more like something Perplexity would be seeding than one that Apple executives would be leaking.

Cloudflare: ‘Perplexity Is Using Stealth, Undeclared Crawlers to Evade Website No-Crawl Directives’ ★

The Cloudflare blog:

We are observing stealth crawling behavior from Perplexity, an AI-powered answer engine. Although Perplexity initially crawls from their declared user agent, when they are presented with a network block, they appear to obscure their crawling identity in an attempt to circumvent the website’s preferences. We see continued evidence that Perplexity is repeatedly modifying their user agent and changing their source ASNs to hide their crawling activity, as well as ignoring — or sometimes failing to even fetch — robots.txt files.

The Internet as we have known it for the past three decades is rapidly changing, but one thing remains constant: it is built on trust. There are clear preferences that crawlers should be transparent, serve a clear purpose, perform a specific activity, and, most importantly, follow website directives and preferences. Based on Perplexity’s observed behavior, which is incompatible with those preferences, we have de-listed them as a verified bot and added heuristics to our managed rules that block this stealth crawling. [...]

Our multiple test domains explicitly prohibited all automated access by specifying in robots.txt and had specific WAF rules that blocked crawling from Perplexity’s public crawlers. We observed that Perplexity uses not only their declared user-agent, but also a generic browser intended to impersonate Google Chrome on macOS when their declared crawler was blocked.

Perplexity has responded, accusing Cloudflare of incompetence and publicity-seeking:

Because Cloudflare has conveniently obfuscated their methodology and declined to answer questions helping our teams understand, we can only narrow this down to two possible explanations.

Cloudflare needed a clever publicity moment and we–their own customer–happened to be a useful name to get them one.

Cloudflare fundamentally misattributed 3-6M daily requests from BrowserBase’s automated browser service to Perplexity, a basic traffic analysis failure that’s particularly embarrassing for a company whose core business is understanding and categorizing web traffic.

Whichever explanation is the truth, the technical errors in Cloudflare’s analysis aren’t just embarrassing — they’re disqualifying. When you misattribute millions of requests, publish completely inaccurate technical diagrams, and demonstrate a fundamental misunderstanding of how modern AI assistants work, you’ve forfeited any claim to expertise in this space.

Perplexity’s response makes it sound like Cloudflare just doesn’t get how leading-edge AI chatbots work, and what users expect of them. But going back to Cloudflare’s post, they specifically cite OpenAI as an exemplar in respecting the directives of website publishers:

When we ran the same test as outlined above with ChatGPT, we found that ChatGPT-User fetched the robots file and stopped crawling when it was disallowed. We did not observe follow-up crawls from any other user agents or third party bots. When we removed the disallow directive from the robots entry, but presented ChatGPT with a block page, they again stopped crawling, and we saw no additional crawl attempts from other user agents. Both of these demonstrate the appropriate response to website owner preferences.

And nothing in Perplexity’s response attempts to explain Cloudflare’s accusation that Perplexity is adopting a false generic user-agent when their own declared user-agents are disallowed. Seems shifty to me.

‘Apple: The First 50 Years’ – New Book by David Pogue, Coming Next Year ★

Coming March 17, 2026:

In time for Apple’s 50th anniversary, “CBS Sunday Morning” correspondent David Pogue tells the iconic company’s entire life story: how it was born, nearly died, was born again under Steve Jobs, and became, under CEO Tim Cook, one of the most valuable companies in the world.

The 600-page book features 360 full-color photos, new facts that correct the record and illuminate Apple’s subversive culture, and 150 fresh interviews with the legendary figures who shaped Apple into what it is today.

Lenovo ThinkBook Plus Gen 6 Has a Rollable Expanding Display ★

Antonio G. Di Benedetto:

Part of me still can’t believe it, but Lenovo did the thing: it took a bonkers concept for a laptop with a rollable screen and built the tech into something you can actually own and use like a normal computer. Except, as conventional as the ThinkBook Plus Gen 6 can be, it’s far from a normal computer. It’s a $3,300 laptop with a screen that expands from 14 inches to 16.7 inches at the push of a button.

Oh, and it’s actually good. Not just good, but very good. I still can’t believe it.

Di Benedetto, as you can see, is enthusiastic for the laptop. I think it’s a clever idea, but this first instance seems pretty compromised:

As with a foldable phone, you can see some creases and ripples in the screen’s lower third — the part that rolls up — especially at oblique angles. If I look closely while working on a bright-white document, I can sometimes make out a faint shadowy strip, but I rarely see it, even when staring at that spot. The motorized screen takes about eight seconds to extend or retract, and it’s no louder than the fans on an average gaming laptop. People right near you in a quiet space will hear it, but even ambient sounds like a TV in the background easily mask the motor.

The New York Post Is Expanding to LA, Launching The California Post Next Year ★

Alexandra Steigrad, reporting for The New York Post:

The new publication will be headquartered in Los Angeles and feature a robust staff of editors, reporters and photographers dedicated to covering news, entertainment, politics, culture, sports and business — all with a distinctly California perspective.

The California Post will be supported by the team in New York providing national and international news. The content will appear in a daily print edition and will have its own dedicated homepage for Californians with stories being published across multiple other platforms, including video, audio and social media.

Who says print is dead?

Apple Q3 2025 Results and Charts ★

Jason Snell, at Six Colors:

On Thursday, Apple reported its third-quarter 2025 fiscal results. Revenue was $94 billion (a fiscal third-quarter record), up 10% versus the year-ago quarter. Mac revenue was up 15%, iPhone revenue up 13%, and Services revenue up 13%. The Wearables/Home/Accessories category was down 9% and iPad revenue down 8%.

Two notes from the analyst call and prepared remarks. First, Apple did, for the first time, acknowledge the risk that Judge Amit Mehta, when he issues his remedies in the US v. Google case, might ban Google from making traffic acquisition cost payments, which would cut off at least $20 billion per year in Apple’s revenue. But while Apple is acknowledging there’s a risk, they’re not giving any hint what they plan to do if that happens. From the call:

Wamsi Mohan, Bank of America: Hi, yes, thank you so much. Tim, I know you said similar growth in Services and that’s predicated with Google payments continuing. Is there any way for us to dimensionalize or maybe just conceptually talk about maybe options if the counter were to happen, if the payments were not allowed in some way? What are some of the things that Apple could do given that it is a significant chunk of profitability?

Tim Cook: Yeah, Wamsi, I don’t really want to speculate on the court ruling and how they would rule and what we would do as a consequence of it.

Wamsi Mohan, Bank of America: OK, I guess we’ll wait for that ruling to come out.

Yes, I guess we will.

Similarly, on AI strategy and which aspects Apple sees as commodities and which it deems as essential and proprietary:

Krish Sankar, TD Cowen: Tim, I’m curious about your thoughts on AI for edge devices. You know, there’s like some people who think that LLM could be a commodity in the future. Do you see a scenario where LLMs become a core part of your iOS, or is the SLM the way to go, and how to think about evolution of edge devices in a futuristic AI world, and is smartphone going to be the choice of device? I’m curious your thoughts on it, broadly speaking.

Tim Cook: The way that we look at AI is that it’s one of the most profound technologies of our lifetime, and I think it will affect all devices in a significant way. What pieces of the chain are commoditized and not commoditized, I wouldn’t want to really talk about today because that gives away some things on our strategy, but I think it’s a good question.

These quarterly calls are better than nothing, but when it comes to anything not in their prepared statements, Apple seldom reveals anything at all.

Sunday, 3 August 2025

Hello Weather ★

My thanks to Hello Weather for sponsoring last week at DF. Regular readers know that I am an inveterate aficionado of weather apps. I’m not really much of a meteorology nerd, but I’m a pedestrian in a city with widely varying seasons. But even more so, I find weather apps to be a true playground for UI design and the presentation of quantitative information. Different weather apps take very different approaches, and I find the differences fascinating.

I first recommended Hello Weather back in 2021, and it’s been in my regular rotation of weather apps ever since. Back at the beginning of summer when they booked the sponsorship for this week, I started using the new version 4.0 (now up to 4.1.4) as my daily driver. I love it. It even has one of my favorite features for hot humid summers — a preference to set “feels like” as the primary temperature display (including in places like widgets). Hello Weather remains, as ever, attractive and useful in its design. And it offers everything you want in an iOS weather app: widgets for the home screen and Lock Screen, a Watch app, notification options (precipitation, severe storms, morning/evening forecast reports), and, wow, a veritable slew of forecast data providers to choose from.

Hello Weather’s privacy story is perfect: they collect zero user data, have no tracking or ads, and their privacy policy is written by humans for humans.

I highly recommend you download Hello Weather and start a 7-day free trial.

How to Leave Substack ★

Simple single-page website with (a) reasons to leave Substack; (b) links to comprehensive step-by-step instructions for how to move to other platforms, such as Ghost, Buttondown, and Beehiiv; and (c) links to several popular publications that moved and are glad they did.

Google Reverses Course, Will Keep goo.gl Links Redirecting ★

Google:

While we previously announced discontinuing support for all goo.gl URLs after August 25, 2025, we’ve adjusted our approach in order to preserve actively used links. We understand these links are embedded in countless documents, videos, posts and more, and we appreciate the input received.

Nine months ago, we redirected URLs that showed no activity in late 2024 to a message specifying that the link would be deactivated in August, and these are the only links targeted to be deactivated. If you get a message that states, “This link will no longer work in the near future”, the link won’t work after August 25 and we recommend transitioning to another URL shortener if you haven’t already.

All other goo.gl links will be preserved and will continue to function as normal. To check if your link will be retained, visit the link today. If your link redirects you without a message, it will continue to work.

Nice!

Saturday, 2 August 2025

Ana Marie Cox on the Shaky Foundation of Substack as a Business ★

Ana Marie Cox, who knows a thing or two about indie publishing and journalism, on her AMC All the Time blog about a month ago:

My take is more dire, because I’m not sure about “savvy and stamina” as the distinguishing characteristics of those who might be able to migrate elsewhere. I think plenty of smart folks might find themselves stuck.

Substack is rickety. It’s as unstable as a SpaceX launch, as overpromised as a Stephen Miller marriage.

Substack does not have a clear future as a newsletter business, I’m not the first to notice that. But it doesn’t have to fail outright to be a disaster. It just has to keep trying to become a life-sized map of the internet: maximum content, maximum churn. The center cannot hold — especially not for newsletters, a format that depends on intimacy and long-standing trust.

The Substack bust will not just take out a few hot-take merchants and media dilettantes. It’s going to take down a lot of working journalists who’ve built modest, sustainable incomes as well as the fragile public sphere we’ve been piecing together in the ashes of Twitter and the twilight of traditional journalism.

Taylor Lorenz’s scoop today on Substack’s Nazi notification oopsie reminded me that I’ve been meaning to link to this post from Cox casting a serious stink eye at Substack’s business. As she says up front, “Let’s set Substack’s ‘Nazi problem’ aside for a moment. What if the bigger issue is being stranded on a collapsing platform ... with a bunch of Nazis?”

Substack pitches itself to would-be independent writers as a thriving platform that’s fundamentally about independent blog publishing and email newsletter distribution. That could be a great business. But it would be a relatively small business compared to Substack’s fund raising (over $100 million so far, and currently looking to raise more) and the implied valuation that fund raising implies (at one point, they were pitching investors that they were worth $1 billion, which is about as realistic as El Gringo Loco Anaranjado’s promises that Mexico will pay for a US border wall).

Ghost is a platform and business that’s actually built for independent writers. So is Buttondown (which Cox uses for her site). Beehiiv too. There’s a whole cottage industry of creator-oriented blog-cum-newsletter platforms. Substack, on the other hand, is a trap. It breaks my heart to see great writers as disparate as Paul Krugman and Michael Chabon set up their ostensibly independent presences on Substack. Writers check in, but — if Substack gets their way — they won’t check out.

Looking at the numbers Cox lays out, Substack’s future looks even worse than I thought. Before they go under, though, their investors will put the screws to them, and Substack will take its heel turn.

Substack Sends Notification Promoting Nazi Blog ★

Ashley Belanger, writing for Ars Technica:

After Substack shocked an unknown number of users by sending a push notification on Monday to check out a Nazi blog featuring a swastika icon, the company quickly apologized for the “error,” tech columnist Taylor Lorenz reported. [...]

Substack has long faced backlash for allowing users to share their “extreme views” on the platform, previously claiming that “censorship (including through demonetizing publications)” doesn’t make “the problem go away — in fact, it makes it worse,” Lorenz noted. But critics who have slammed Substack’s rationale revived their concerns this week, with some accusing Substack of promoting extreme content through features like their push alerts and “rising” lists, which flag popular newsletters and currently also include Nazi blogs.

Gizmodo reports:

The publication in question, NatSocToday, describes itself as a “National Socialist weekly newsletter featuring opinions and news important to the National Socialist and White Nationalist Community.” The newsletter’s image header is a Nazi flag, and its latest post, as of Wednesday, was an article that includes the sentence: “We demand the return of all territory currently occupied by jews and non-Whites in historically White homelands.” It does not appear to be a particularly popular blog, and currently has fewer than a thousand subscribers.

A mantra of “we host all views, but don’t promote or endorse all views” doesn’t hold much water when you promote a blog whose logo is a straight-up Nazi swastika.

Friday, 1 August 2025

The Talk Show: ‘Michigan-Starred Fine Dining’ ★

Special guest Louie Mantia joins the show to talk about Liquid Glass, the various OS 26 updates, and the worrisome state of Apple’s UI design overall. Also: sandwiches.

Sponsored by:

Quip: A supercharged clipboard manager and text expander for Mac, iPhone, and iPad.
Notion: The best AI tool for work, with your notes, docs, and projects in one space.
Squarespace: Save 10% off your first purchase of a website or domain using code talkshow.