Published
Weekend Reading — Cloud outages will continue until morale improves
Tech Stuff
I am sorry, but everyone is getting syntax highlighting wrong I'm using Alabaster to see how I like it (PS it's an extension, not a theme):
The problem with that is, if everything is highlighted, nothing stands out. Your eye adapts and considers it a new norm: everything is bright and shiny, and instead of getting separated, it all blends together.
Always test your apps IN PERSON with real people by looking over their shoulder.
You will learn more from one person than from 100 TestFlight users.
Writing it down here because I keep forgetting it.
The Pragmatic Engineer 2025 Survey: What’s in your tech stack? Part 3
Which tools do software engineers use for observability, oncall tooling, feature flags, frontend & mobile work, and for developer tooling? Results from our survey, based on 3,000+ responses by readers
These massive "cloud" outages will continue to happen, inevitably, because they are "normal accidents": Any system that:
- is complex,
- is tightly coupled, and
- has catastrophic potential…can be expected to experience catastrophic failures. AWS, Google Cloud, Microsoft Azure check all these boxes.
Fun fact: Internet was specifically designed to not be like that. It was designed to be loosely coupled.
But that's not as good for maximizing shareholder value!
The best Chrome browser isn’t Chrome, it’s Helium
Helium offers an alternative to Google Chrome for Mac, Windows, and Linux users who prefer not to have AI Mode and Gemini enabled in the browser and prioritize privacy over Google's data collection practices.
I was today years old when I learned the phrase "swivel chair integration" (as in, get info from one system then swivel your chair around to type that same information into a different system instead of the two systems actually being integrated at, say, an API level).
What a delightful phrase. How sad that it is still a thing, and the phrase isn't as outdated as manual telephony switchboards...
Locally AI Run Llama, Gemma, Qwen, DeepSeek, and more on your iPhone or iPad. Offline and private.
Eye for Design
Error handling - UX design patterns "The best way of helping users fix an error, is to prevent it in the first place" but if you do have to show error messages, here's a quick intro with some UX inspiration for nicer error handling.
Free SVG icon for web developers These icons are for developers so you get icons like cup of coffee, PHP, lollipop, and other delights.
GenAI Image Showdown Comparing 14 models, including OpenAI 4o, Gemini 2.5, Imagen 4, and Midjourney v7.
Business Side
Rotate 180° (via Ernie Smith)
Canva just flipped 40 years of design-business logic on its head, and if they pull this off, they look like geniuses. If they don’t, hopefully Affinity doesn’t get lost in the shuffle.
This is an extremely smart rebrand. Grammarly sounds like a product that can be replaced by a ChatGPT prompt while Superhuman sounds like a product that gives you superpowers.The acquisition was worth it just for that alone.
Paid Armies of 'Clippers' Boost Internet Stars Like MrBeast
“People used to buy commercials on TV, billboards, radio time slots,” said Anthony Fujiwara, the 23-year-old founder of Clipping, a marketing service that he named after this new social-media-marketing technique. “Clipping is that for the modern era. It’s buying space and time on people’s phones while they scroll.”
Machine Intelligence
Silicon Valley's AI agents can't schedule meetings but promise to replace workers by 2027 (via mtmail)
This is the maddening reality of artificial intelligence in 2025: the same technology that can diagnose rare diseases from medical scans can't handle the calendar conflict between British Summer Time and Eastern Daylight Time. OpenAI's GPT-4o—supposedly the apex of machine intelligence—fails at basic office tasks 91.4% of the time.
Lost in the Middle: How Language Models Use Long Contexts I can't tell if they're talking about LLMs or how everyday people read and remember text:
In particular, we observe that performance is often highest when relevant information occurs at the beginning or end of the input context, and significantly degrades when models must access relevant information in the middle of long contexts, even for explicitly long-context models.
All fail Caesar: students at eight schools taught wrong topic for final year 12 history exam AI may hallucinate but human teachers always get the facts correct, except when they don't:
Year 12 students from at least eight schools in Queensland were taught the wrong topic for their final history exams and authorities are now checking with 172 schools to see if any more were affected.
An Opinionated Guide to Using AI Right Now In my opinion this is a really good article and if you're using AI right now I recommend reading it.
Insecurity
"disable your ad blocker" is the "take off the condom" of corporate media
📣THREAD: It’s surprising to me that so many people were surprised to learn that Signal runs partly on AWS (something we can do because we use encryption to make sure no one but you–not AWS, not Signal, not anyone–can access your comms).
It’s also concerning. 1/
Everything Else
LinkedIn Wisdom Generator "Amplifying opportunity is a choice, not a bandwidth"
My favorite homemade cleaning recipe is one cup of the cheapest vodka you can get, and the juice of one or two limes. Drink it and repeat until you no longer care about cleaning
So many problems around the house can be solved by simply learning to not look at that part of the house
When TVs were furniture and screens were tiny 90 years and walking backwards.
If you're publishing a magazine which is a consumer guide to electric cars and you're not calling it WHAT EVS then you're wasting everybody's fucking time.
I don't want to chase my dreams. I want to hire a private investigator to follow my dreams around town and gather information. Then when the time is right I'll confront my dreams with what I know of it and blackmail it into giving me what I want.
Over 1,000 items looted in California museum heist Le Monde would like to inform you there was an epic musuem heist that was ongoing in the Not-French(tm) town of Oakland Californica.
The mathematician John Conway coined the delightful term "probviously" for situations where the specific problem is very hard to solve, but reasoning about the typical behavior of similar problems makes the answer seem obvious.
Once again had to download a stupid app and spend ten minutes by the side of the road faffing with my phone to charge the car. This is such a huge barrier to entry for so many people, I do not know why you cannot just plug your car in and swipe a bank card, it is ridiculous. No wonder people stick to petrol cars, at petrol stations you can just swipe your card and fill up. It's inexcusable that it isn't the same experience for electric, there's no good reason for it.
The World's FirstNon-SmartAI Pendant
Introducing the NotFriend - the ultimate in anti-technology fashion. This literal plastic circle does absolutely nothing, but looks great doing it.
Fun fact: the coffee gods are real. Whenever you see coffee nerds finagling with 10 different tools and accessories throughout their coffee prep routine they're not actually doing so because they have a solid empirical reason to believe it improves their coffee. It's all an occult ritual designed to please the gods before pulling the shot. We don't actually know why or how any of this works but it seems to make the gods happy.
“Hi I’m Doctor Smith. I understand you took a fall earlier Miss Cassidy. Do you have any pain anywhere?”
“Yes Doctor, my arm.”
“Ok, and if 10 is the worst pain you’ve ever felt and 0 is no pain at all, how would you rate the pain?”
“Erm, somewhere between 6-7.”
“6-7 you say? Nurse, get me my nephew on FaceTime please.”
Build-a-Bear charging me $10 for “e-shipping” for the electronic gift card I’m required to purchase as a deposit to book my son’s birthday party. How to maximize revenue from consumers.
Italian Espresso Isn't What You Think: A Data-Driven guide
My takeaway from this trip is that we can certainly still learn some things from the institution that is Italian espresso. Not by copying their coffee, but by copying their thinking: start with the experience you want to create, then engineer everything else around it.
A group of scientists set out to study quick learners. Then they discovered they don't exist
“Students are starting in different places and ending in different places,” said Ken Koedinger, a cognitive psychologist and director of Carnegie Mellon’s LearnLab, where this research was conducted. “But they’re making progress at the same rates.”
A 6-year research project found a surprisingly simple route to happiness
In one study, Kashdan had patients with social anxiety disorder describe what they saw as their purpose in life and then had them monitor their efforts toward this purpose every day. After two weeks, they exhibited higher self-esteem, sense of meaning and positive emotions. Days when they reported significant progress toward life purpose were also days with significant boosts in well-being.
