*️⃣ Galvanized
On June 30th, 2025, I started working at Stainless as a part-time contractor. A couple months later, on August 21st, I joined the team full-time.
Today, on May 18th, Stainless announced they've been acquired by Anthropic. See also: Anthropic's announcement.
Given how I feel about generative AI, I declined the opportunity to join Anthropic. I wish all of my former colleagues the very best regardless of where they landed, and I'm looking forward to hearing about their new adventures and endeavors.
Stainless was a truly great place to work. Alex Rattray was the best founder anyone could ask for, and he assembled an incredible team. Everyone there—every single person—was incredibly capable and incredibly kind. I'm honored, humbled, and privileged to have been a small part of that team and the work we did.
As for what's next for me, I'm looking forward to taking a break. We've been renting and moving around for many, many years, but early this year we finally bought our we're-going-to-be-here-for-decades house. I'm really looking forward to taking some time to make it our home. As I arrange furniture and turn screws I'll have a background process running to figure out what's next career-wise.
I'm also looking forward to blogging more. With my newly refined blogging setup, its lower friction authoring process, and more free time, expect the frequency of my posts to increase.
My sincere and heartfelt thanks go out to Alex and every member of the Stainless team. It was a lot of fun to work on a small team at a startup for a little while, and I'm sure it's quite possible some of us will work together again in the future.
🔣 My Favorite Regular Expressions Tool: regex101
Whenever I need to do anything non-trivial with text I'm so grateful regex101 exists. It's the best tool I've found for creating, understanding, and debugging regular expressions by a mile.
I made heavy use of it when rewriting this site's CMS again, and a couple of times it was the only thing keeping my brain from leaking out of my ears.
🛳️ A Great Hantavirus Overview
Violet Blue has a great overview of the Hantavirus situation on Threat Model:
I devoted a significant space in this week’s pandemic roundup to the hantavirus outbreak. That’s because I’m not seeing a well-rounded update putting the whole picture in focus – and I’m seeing a wild amount of misinformation get reported by every kind of news outlet. Even contradicting headlines from mainstream outlets: for instance, Forbes ran a headline yesterday saying there were no positive cases in the US, coming up in searches right next to headlines about the positive case in Nebraska. It’s like having newsgathering whiplash. It’s all the more important that I take the time to check sources and claims, compare reports, and look at timelines. That way I can give you an accurate snapshot – and know for myself what the hell is going on.
Violet's coverage of the ongoing pandemic, and other threats to health and tech, have been an invaluable resource for a long time, and she makes it all available for free thanks to her paid supporters, including yours truly. If you find her work valuable, consider becoming a paid member.
Note that I'm not affiliated with Violet Blue other than being a member of her Patreon. This isn't a paid endorsement or anything like that (which you won't find on this site in any form), I just want to see people doing public good get compensated for it.
📚 Jenny Volvovski
Jenny Volvovski designs incredible book covers. Delightful and inspiring.
See also: Also, her design studio where she does more great work.
🚀 2026 Relaunch
Last year I wrote a content management system from scratch just for this site. It was a lot of fun, and it gave me the opportunity to do things exactly the way I wanted to do them, which was liberating.
When I wrote that CMS, one of my primary goals was simplicity. I mostly succeeded, but the siren song of complexity, feature creep, and wouldn't-it-be-cool-if is hard to ignore. Thus, complexity crept in and introduced a little too much friction in the authoring process.
Don't get me wrong; that CMS served me well and gave me the chance to explore some new and interesting ideas. Some things, however, just don't pan out the way I expected.
I thought about trying to extract the complexity from that CMS, but after a half-hearted attempt to do so I realized it would be better to start from scratch again. (Well, mostly from scratch. I copied and pasted a lot of good stuff from the old one.)
So, welcome to the 2026 relaunch of my site! (If you're reading this in a feed reader, check it out in your web browser to see the visual changes!)
What's different
This version of the CMS omits several things the old one had. Tags are completely gone, and posts no longer have unique colors or summaries. Instead, I'm only putting an emoji in the title and using that to derive a fun visual appearance for each post.
There's also no longer a distinction between link posts and regular posts. All posts are now just regular posts. I've gone through my previous posts and made adjustments to accommodate those changes as needed.
Another big change: there's not currently a light mode. I might implement one in the future, but for now I really like having a single visual appearance to focus on rather than one with dual modes.
All of those changes eliminate a lot of friction from the writing and publishing process, which is exactly what I wanted. I found it was sometimes a chore to post a quick link or thought, and that meant this site wasn't serving its purpose.
Now I have much less to think about when writing a post. The only requirements are a Markdown file with a title that contains an emoji. I don't have to think about what tags to use, I don't have choose a color, I don't have to decide between a link post or a regular post. I just create a Markdown file and write.
I'd love to hear your thoughts on these changes. You can find the link to email me on my new about page.
Now, time to tackle that backlog of things I've been wanting to write about!
🦠 Ultrasound waves rupture COVID-19 and flu viruses without damaging cells
Maria Fernanda Ziegler:
The discovery surprised the researchers because it contradicts classical physics theories, as the wavelength of ultrasound is much longer than the size of the virus. In theory, this difference in size would prevent interaction.
"The phenomenon is entirely geometric. Spherical particles, such as many enveloped viruses, absorb ultrasound wave energy more effectively. It's that accumulation of energy inside the particle that causes changes in the structure of the viral envelope until it ruptures. Therefore, if viruses were triangular or square, they wouldn't undergo the same 'popcorn effect' of acoustic resonance," Bruno explains.
He also points out that since the process depends strictly on the shape of the viral particle and not on genetic mutations, variants such as those observed during the pandemic (omicron and delta, for example) do not affect the effectiveness of the technique.
That's cool as hell.
Hopefully this pans out into a workable treatment (which will probably be years from now, but I'll take what I can get at this point).
💥 “They Would Never Use the Death Star on Us”: Alderaan Residents Reflect on Their Support for the Empire as a Large Imperial Installation Enters the System
Jack Loftus over at McSweeny's:
Lesser of two evils. The Senate was ineffective, and the liberal Jedi were out of touch. The Emperor said he’d cut through all that. And he did—sometimes literally. You have to give him that. Things moved. Maybe a little too much moving right now, with the Death Star repositioning every few minutes to maintain a firing solution on our planet, but still.
Yeah.
🔊 Hostile Volume
Just set the volume to 25%. Just fold in the cheese.
🏎️ International Chess Federation Adds Race Car Piece
The Onion:
In all officially sanctioned matches played from today forward, the pawn immediately in front of a player’s king will be replaced with a sick little hot rod that can move any number of squares horizontally, vertically, or in a circle like it’s doing donuts[.]
Delightful piece. Do pay them a visit to learn all the new rules.
🎲 Claude mixes up who said what, and that's not OK
Gareth Dwyer, in a post about LLMs not keeping track of who said what:
Yes, of course AI has risks and can behave unpredictably, but after using it for months you get a ‘feel’ for what kind of mistakes it makes, when to watch it more closely, when to give it more permissions or a longer leash.
Sure, you can get a “feel” for what kind of mistakes LLMs make, but that feeling means absolutely nothing and cannot be trusted. Randomness is literally built in. The temperature setting of an LLM determines how random the token selection process will be, and every token selection is a fork in the road for the rest of the output.
This kind of widespread misunderstanding about how LLMs work terrifies me. To be clear, I’m not picking on Gareth specifically—this is just a good example of the misplaced trust in LLMs I see running rampant these days. These systems are intentionally designed to be nondeterministic, with randomness as a key component of how they work. Thinking you can get a “feel” for when you can trust them is like getting a “feel” for how the dice are rolling or a “feel” for how the cards are being shuffled.