John’s Blog

My personal journal and blog. Subscribe via RSS


June 25, 2025

Anthropic and Fair Use

Olivia Sophie Rafferty, for ai fray:

The Northern District of California has granted a summary judgment for Anthropic that the training use of the copyrighted books and the print-to-digital format change were both “fair use” […]

However, the court also found that the pirated library copies that Anthropic collected could not be deemed as training copies, and therefore, the use of this material was not “fair”. […]

This is a mixed ruling on fair use – a loss for both copyright holders and Anthropic, but potentially a big win for AI platforms in general. And, if upheld, the order would mean that AI firms using copyrighted material to train their LLMs may be allowed in the future. The only exception to this would be if the material has been pirated.

The mixed ruling in this case is potentially very interesting for the larger AI and LLM-based industry. The court here is establishing some precedent that training an LLM model is considered ‘transformative’ because it uses existing copyrighted works to create new outputs. This ruling sure seems to validate the approach of training LLMs on large datasets of copyrighted materials as acceptable, in the eyes of the court at least.

On the other hand, ‘pirating’ a bunch of books and content is obviously not fair use, and Anthropic will be on the hook for those damages in a future trial.

June 24, 2025

No Io

Hayden Field, for The Verge:

OpenAI has scrubbed mentions of io, the hardware startup co-founded by famous Apple designer Jony Ive, from its website and social media channels. The sudden change closely follows their recent announcement of OpenAI’s nearly $6.5 billion acquisition and plans to create dedicated AI hardware.

OpenAI tells The Verge the deal is still happening, but it scrubbed mentions due to a trademark lawsuit from Iyo, the hearing device startup spun out of Google’s moonshot factory.

Turns out that one of the hardest things in computer science still applies to our AI-driven world: naming things.

And really: “io” is such a common phrase in computing, it shouldn’t be able to be trademarked at all.

June 13, 2025

The Newspaper That Hired ChatGPT

Matteo Wong, writing for The Atlantic, interviewed Claudio Cerasa, editor of the Italian newspaper Il Foglio about his use of AI tools and how they are evolving his paper:

With so many obvious pitfalls to using AI, I wanted to speak with Cerasa to understand more about his experiment. Over Zoom, he painted an unsettling, if optimistic, portrait of his experience with AI in journalism. Sure, the technology is flawed. It’s prone to fabrications; his staff has caught plenty of them, and has been taken to task for publishing some of those errors. But when used correctly, it writes well—at times more naturally, Cerasa told me, than even his human staff.

Still, there are limits. “Anyone who tries to use artificial intelligence to replace human intelligence ends up failing,” he told me when I asked about the “Heat Index” disaster. “AI is meant to integrate, not replace.” The technology can benefit journalism, he said, “only if it’s treated like a new colleague—one that needs to be looked after.”

The problem, perhaps, stems from using AI to substitute rather than augment. In journalism, “anyone who thinks AI is a way to save money is getting it wrong,” Cerasa said.

June 12, 2025

Giving Dia a Spin

Dia, the new AI-focused browser by The Browser Company, is now available for use by Arc members. I'm not even sure what an "Arc member" is, but I've been a heavy Arc user for a few years now so I'm giving it a spin.

A screenshot of the Dia browser summarizing an article about the new F1 movie.
Dia summarizing an open tab

The interface is super clean and polished. It’s very tastefully designed. I haven’t quite figured out what I’ll use the AI chat sidebar for on a daily basis, but I’m giving it a shot.

The onboarding process was very useful as well—a rarity for browsers! Since I’m an Arc user, Dia imported all of my bookmarks, site preferences, and even my cookies and browsing data, so I haven’t had to sign-in to anything new and can just continue along my way like nothing has changed. Fancy!

I asked the chat interface to summarize the value proposition of Dia and its core features. Here’s what it said about itself:

Dia is a new AI-first browser from The Browser Company, designed to put artificial intelligence at the center of your web experience. Unlike traditional browsers, Dia features a deeply integrated AI assistant that’s always available in a sidebar, ready to help you summarize pages, answer questions, draft emails, or compare information across multiple tabs. Its minimalist, Chrome-like interface makes it approachable for everyday users, not just power users.

Dia streamlines everyday web tasks by making AI assistance seamless and ever-present. Whether you’re researching, writing, shopping, or planning, Dia reduces busywork and context-switching, helping you get more done in less time. Its focus on simplicity, personalization, and privacy aims to make the browser feel less like a tool and more like a helpful partner—one that adapts to your needs and grows smarter the more you use it.

June 11, 2025

Hiding Under Liquid Glass

M.G. Siegler:

Ever since the rise of Android to challenge the iPhone all those years ago, Apple and Google have stood atop the tech landscape when it comes to their big developer-focused events. And their temporal proximity to each other each calendar year have always highlighted the key similarities and differences. But never have they been more different than they were this year. Google I/O was AI, AI, AI, AI, AI, AI. Apple’s WWDC was UI, UI, UI, UI, UI, UI.

One felt like a company flexing. The other felt like a company hiding.

“Hiding” is a strong term. I think this was Apple focusing on what it does better than any other company out there and leaning into its strengths. And the same for Google. The dichotomy of these two conferences makes me wish the two companies got along better and could work together more.

June 10, 2025

Liquid Glass

This portion of the WWDC keynote yesterday was very tight and well presented. The design team at Apple is excellent at this type of update. Much like iOS 7, it instantly makes everything else on our devices feel outdated and stale. It will be interesting to see how quickly the rest of the industry attempts to copy this design language.

June 9, 2025

iPadOS 26

Apple announced some great changes for the iPad today that have me very excited for the fall releases.

Finally, window controls:

iPadOS 26 introduces powerful new features that help users work with, control, organize, and switch between app windows — all while maintaining the immediacy and simplicity that iPad users expect. The new windowing system lets users fluidly resize app windows, place them exactly where they want, and open even more windows at once.

And a files app that works the way it should:

iPadOS 26 introduces powerful new ways to manage, access, and edit files. An enhanced Files app with an updated List view allows users to see more of their document details in resizable columns and collapsible folders. To make folders easier to identify at a glance, Files also offers folder customization options that include custom colors, icons, and emoji that sync across devices. For quicker access, users can now drag any folder from the Files app right into the Dock. Additionally, users can set a default app for opening specific files or file types.

And a menu bar:

With a new menu bar, users can access the commands available in an app with a simple swipe down from the top of the display, or by moving their cursor to the top. Users can quickly find a specific feature or related tips in an app by using search in the menu bar. Additionally, developers can now customize the menu bar in their own apps.

All of these feature have existed on the Mac for decades. These are the iPad updates that we should have seen years ago, but I’m so glad they are finally coming.

June 5, 2025

My AI Skeptic Friends Are All Nuts

Linked to in the aforementioned post by Manton, this post by Thomas Ptacek for Fly.io hits home and deserves a callout:

Some of the smartest people I know share a bone-deep belief that AI is a fad — the next iteration of NFT mania. I’ve been reluctant to push back on them, because, well, they’re smarter than me. But their arguments are unserious, and worth confronting. Extraordinarily talented people are doing work that LLMs already do better, out of spite. […]

Professional software developers are in the business of solving practical problems for people with code. We are not, in our day jobs, artisans. Steve Jobs was wrong: we do not need to carve the unseen feet in the sculpture. Nobody cares if the logic board traces are pleasingly routed. If anything we build endures, it won’t be because the codebase was beautiful.

Besides, that’s not really what happens. If you’re taking time carefully golfing functions down into graceful, fluent, minimal functional expressions, alarm bells should ring. You’re yak-shaving. The real work has depleted your focus. You’re not building: you’re self-soothing.

Which, wait for it, is something LLMs are good for. They devour schlep, and clear a path to the important stuff, where your judgement and values really matter.

Emphasis above is mine. This whole post is so brilliantly written, it’s worth a read and I could quote the whole thing.

June 5, 2025

AI Distrust and Forward Progress

Manton Reece:

The more I think deeply about AI, the more I reflect on humanity and creativity and what our purpose here might be.

I understand feeling distrust for AI on principle. I’ve read so many blog posts from people who have various reasons for wishing AI didn’t work the way it does, didn’t use as much energy, didn’t crawl the web without permission, didn’t put people out of work, didn’t upend education, and so on.

For me, now that I’ve seen AI, I can’t un-see it. I can’t go about my life as if nothing has changed. In a world where machines are smarter than we are, what should we work on? Everyone will find value and happiness in different ways.

A well articulated post. My thoughts exactly here. I especially like this conclusion:

My gut feeling is that for the folks who do not change anything in response to AI, pretending that AI doesn’t exist, they will increasingly be unhappy. Not today, but eventually. Despite all the hype, the changes will creep up on us slowly over several years.

June 5, 2025

Sky

Sky is a new app for the Mac created by the team that previously created the great Workflow app, which later became Shortcuts.

Now, with AI, you can chat with computers the same way you talk to people: natural language. We’re on a journey to bring that capability to everything you can do on your computer.

Introducing Sky, natural computing for the Macintosh.

Sky floats over what you’re doing so AI is always at your fingertips. Whether you’re chatting, writing, planning, or coding, Sky understands what’s on your screen and can take action using your apps.

Federico Viticci has a first look over at MacStories:

What sets Sky apart from anything I’ve tried or seen on macOS to date is that it uses LLMs to understand which windows are open on your Mac, what’s inside them, and what actions you can perform based on those apps’ contents. It’s a lofty goal and, at a high level, it’s predicated upon two core concepts. […]

At the same time, Sky allows power users to make their own tools that combine custom LLM prompts with actions powered by Shortcuts, shell scripts, AppleScript, custom instructions, and, down the road, even MCP. All of these custom tools become native features of Sky that can be invoked and mixed with natural language.

We’re at the beginning of a whole new frontier of apps like this powered by LLMs. Can’t wait to give this one a spin.

May 28, 2025

Claude 4

Anthropic:

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.

The intro video “A day with Claude” is a cool picture of how some of the integrations work.

Claude Opus 4 also dramatically outperforms all previous models on memory capabilities. When developers build applications that provide Claude local file access, Opus 4 becomes skilled at creating and maintaining ‘memory files’ to store key information. This unlocks better long-term task awareness, coherence, and performance on agent tasks[.]

Simon Willison has an excellent writeup and analysis on the prompts and details of Claude 4.

May 28, 2025

Texas Requires Age Verification for App Stores

Stephen Nellis, for Reuters:

Texas Governor Greg Abbott on Tuesday signed into law a bill requiring [Apple] and [Google] to verify the age of users of their app stores, putting the second-most-populous U.S. state at the center of a debate over whether and how to regulate smartphone use by children and teenagers.

The law, effective on January 1, requires parental consent to download apps or make in-app purchases for users aged below 18. Utah was the first U.S. state to pass a similar law earlier this year, and U.S. lawmakers have also introduced a federal bill.

It was reported last week that Tim Cook called Governor Abbott to try and prevent this from passing.

“The problem is that self-regulation in the digital marketplace has failed, where app stores have just prioritized the profit over safety and rights of children and families,” Casey Stefanski, executive director for the Digital Childhood Alliance, told Reuters.

Apple and Google opposed the Texas bill, saying it imposes blanket requirements to share age data with all apps, even when those apps are uncontroversial.

“If enacted, app marketplaces will be required to collect and keep sensitive personal identifying information for every Texan who wants to download an app, even if it’s an app that simply provides weather updates or sports scores,” Apple said in a statement.

This is yet another case where Apple, and Google, have been reluctant (unwilling?) to do the right thing on their own and therefore government regulation is required.

Screen Time is Apple’s official framework and toolkit for managing child usage and limits on iOS. It doesn’t work. The restrictions are easy to bypass, especially for a modern tech-savvy teen, and the controls it gives parents are wholly insufficient. My own kids shock me on a regular basis for their clever ways to get around the limitations we place on their phones with Screen Time.

If Apple were serious about protecting access to sensitive content or protecting personal identifying information, they would fix the core issues before they become such a problem that lawmakers need to step in and regulate.

May 28, 2025

Arc and Dia

Josh Miller, co-founder of The Browser Company, has posted a thoughtful writeup about the decisions he and his team made to move on from the Arc browser.

After a couple of years of building and shipping Arc, we started running into something we called the “novelty tax” problem. A lot of people loved Arc — if you’re here you might just be one of them — and we’d benefitted from consistent, organic growth since basically Day One. But for most people, Arc was simply too different, with too many new things to learn, for too little reward.

To get specific: D1 retention was strong — those who stuck around after a few days were fanatics — but our metrics were more like a highly specialized professional tool (like a video editor) than a mass-market consumer product, which we aspired to be closer to.

I wouldn’t call myself a “fanatic” but I do really like the Arc browser model and have been using it as my daily driver for several years now. It’s definitely a “highly specialized professional tool”, and I can completely understand why it’s too complicated and different for the mass market.

Switching browsers is a big ask. And the small things we loved about Arc — features you and other members appreciated — either weren’t enough on their own or were too hard for most people to pick up. By contrast, core features in Dia, like chatting with tabs and personalization features, are used by 40% and 37% of DAUs respectively. This is the kind of clarity and immediate value we’re working toward.

Good on the team for properly measuring the usage of these features and understanding what the market is using, rather than what they are asking for.

Miller concludes with some notes about Dia, the new product the company is working on:

I want to end by being frank with you: Dia is not really a reaction to Arc and its shortcomings. No. Imagine writing an essay justifying why you were moving on from your candle business at the dawn of electric light. Electric intelligence is here — and it would be naive of us to pretend it doesn’t fundamentally change the kind of product we need to build to meet the moment. […]

Dia may not be your style. It may not land right away. But this is still us. Being ourselves. Building the kind of thing we’d want to use. Fully aware that we might be wrong. But doing it anyway.

I’m excited to give Dia a spin. I wish I had an .edu email account but, alas, I’ll wait for my turn.

May 23, 2025

OpenAI and io

The big news this week in the tech world was OpenAI’s acquisition of Jony Ive’s hardware startup, io, for $6.5 billion worth of stock. I was trying to catch up on Google IO from this week, but everything there seems overshadowed by this news.

Here’s the announcement video from OpenAI, which is a bit over the top, but I confess it had me mesmerized for the entire nine minutes of this intro. This is a clear and profound statement to the rest of the industry.

John Gruber sums up my feeling, as usual, perfectly:

This is just a vibes teaser, but the vibe is a shot across the bow. It conveys grand ambition, but without pretension. To say I’m keen to get my hands on what they’re making is an understatement.

Mark Gurman and Shirin Ghaffery for Bloomberg have an interesting note here:

While Ive and LoveFrom will remain independent, they will take over design for all of OpenAI, including its software. Altman said his first conversations with Ive weren’t about hardware, but rather about how to improve the interface of ChatGPT.

“Design for all of OpenAI” is a huge deal.

🍿

May 16, 2025

Codex

OpenAI, announcing Codex:

Today we’re launching a research preview of Codex: a cloud-based software engineering agent that can work on many tasks in parallel. Codex can perform tasks for you such as writing features, answering questions about your codebase, fixing bugs, and proposing pull requests for review; each task runs in its own cloud sandbox environment, preloaded with your repository. […]

Codex can read and edit files, as well as run commands including test harnesses, linters, and type checkers. Task completion typically takes between 1 and 30 minutes, depending on complexity, and you can monitor Codex’s progress in real time.

Once Codex completes a task, it commits its changes in its environment. Codex provides verifiable evidence of its actions through citations of terminal logs and test outputs, allowing you to trace each step taken during task completion. You can then review the results, request further revisions, open a GitHub pull request, or directly integrate the changes into your local environment.

Impressive stuff, as usual, from OpenAI.