During the most recent episode of The Talk Show, Jason Snell brought up a weird issue that I started running into last year. On my Mac, sometimes I’d drag an image out of a web page in Safari, and I’d get an image in WebP format. Sometimes I wouldn’t care. But usually when I download an image like that, it’s because I want to publish (or merely host my own copy of) that image on Daring Fireball. And I don’t publish WebP images — I prefer PNG and JPEG for compatibility.

What made it weird is when I’d view source on the original webpage, the original image was usually in PNG or JPEG format. If I opened the image in a new tab — just the image — I’d get it in PNG or JPEG format. But when I’d download it by dragging out of the original webpage, I’d get a WebP. This was a total WTF for me.

I turned to my friend Jeff Johnson, author of, among other things, the excellent Safari extension StopTheMadness. Not only was Johnson able to explain what was going on, he actually made a new Safari extension called ChangeTheHeaders that fixed the problem for me. Johnson, announcing ChangeTheHeaders last year:

After some investigation, I discovered that the difference was the Accept HTTP request header, which specifies what types of response the web browser will accept. Safari’s default Accept header for images is this:

Accept: image/webp,image/avif,image/jxl,image/heic,image/heic-sequence,video/*;q=0.8,image/png,image/svg+xml,image/*;q=0.8,*/*;q=0.5

Although image/webp appears first in the list, the order actually doesn’t matter. The quality value, specified by the ;q= suffix, determines the ranking of types. The range of values is 0 to 1, with 1 as the default value if none is specified. Thus, image/webp and image/png have equal precedence, equal quality value 1, leaving it up to the web site to decide which image type to serve. In this case, the web site decided to serve a WebP image, despite the fact that the image URL has a .png suffix. In a URL, unlike in a file path, the “file extension”, if one exists, is largely meaningless. A very simple web server will directly match a URL with a local file path, but a more complex web server can do almost anything it wants with a URL.

This was driving me nuts. Thanks to Johnson, I now understand why it was happening, and I had a simple set-it-and-forget-it tool to fix it. Johnson writes:

What can you do with ChangeTheHeaders? I suspect the biggest selling point will be to spoof the User-Agent. The extension allows you to customize your User-Agent by URL domain. For example, you can make Safari pretend that it’s Chrome on Google web apps that give special treatment to Chrome. You can also customize the Accept-Language header if you don’t like the default language handling of some website, such as YouTube.

Here’s the custom rule I applied a year ago, when I first installed ChangeTheHeaders (screenshot):

Header: Accept
Value: image/avif,image/jxl,image/heic,image/heic-sequence,video/*;q=0.8,image/png,image/svg+xml,image/*;q=0.8,*/*;q=0.5
URL Domains: «leave blank for all domains»
URL Filter: «leave blank for all URLs»
Resource Types: image

I haven’t seen a single WebP since.

ChangeTheHeaders works everywhere Safari does — Mac, iPhone, iPad, Vision Pro — and you can get it for just $7 on the App Store.

Jesper, writing at Take:

My hope is that Macintosh is not just one of these empires that was at the height of its power and then disintegrated because of warring factions, satiated and uncurious rulers, and droughts for which no one was prepared, ruining crops no one realized were essential for survival.

My hope is that there remains a primordial spark, a glimpse of genius, to rediscover, to reconnect to — to serve not annual trends or constant phonification, but the needs of the user to use the computer as a tool to get something done.

Julien Khaleghy, CEO of SerpApi:

Google thinks it owns the internet. That’s the subtext of its lawsuit against SerpApi, the quiet part that it’s suddenly decided to shout out loud. The problem is, no one owns the internet. And the law makes that clear.

In January, we promised that we would fight this lawsuit to protect our business model and the researchers and innovators who depend on our technology. Today, Friday, February 20, 2026, we’re following through with a motion to dismiss Google’s complaint. While this is just one step in what could be a long and costly legal process, I want to explain why we’re confident in our position.

Is Google hurting itself in its confusion? Google is the largest scraper in the world. Google’s entire business began with a web crawler that visited every publicly accessible page on the internet, copied the content, indexed it, and served it back to users. It did this without distinguishing between copyrighted and non-copyrighted material, and it did this without asking permission. Now Google is in federal court claiming that our scraping is illegal.

I’ve come around on SerpApi in the last few months. My initial take was that it surely must be illegal for a company to scrape Google’s search results and offer access to that data as an API. But I’ve come around to the argument that what SerpApi is doing to obtain Google search results is, well, exactly how Google scrapes the rest of the entire web to build its search index. It’s all just scraping publicly accessible web pages.

This December piece by Mike Masnick at Techdirt is what began to change my mind:

Look, SerpApi’s behavior is sketchy. Spoofing user agents, rotating IPs to look like legitimate users, solving CAPTCHAs programmatically — Google’s complaint paints a picture of a company actively working to evade detection. But the legal theory Google is deploying to stop them threatens something far bigger than one shady scraper.

Google’s entire business is built on scraping as much of the web as possible without first asking permission. The fact that they now want to invoke DMCA 1201 — one of the most consistently abused provisions in copyright law — to stop others from scraping them exposes the underlying problem with these licensing-era arguments: they’re attempts to pull up the ladder after you’ve climbed it.

Just from a straight up perception standpoint, it looks bad.

Apple (MacRumors, Hacker News): With M4, iPad Air is up to 30 percent faster than iPad Air with M3, and up to 2.3x faster than iPad Air with M1. The new iPad Air also features the latest in Apple silicon connectivity chips, N1 and C1X, delivering fast wireless and cellular connections — and support for […]

Apple (MacRumors, Hacker News): At the heart of iPhone 17e is the latest-generation A19, which delivers exceptional performance for everything users do. iPhone 17e also features C1X, the latest-generation cellular modem designed by Apple, which is up to 2x faster than C1 in iPhone 16e. The 48MP Fusion camera captures stunning photos, including next-generation portraits, […]