Published
Weekend Reading — The background noise of the AI bubble
Fake Metro Trains "Due to ongoing delays with the delivery of our new Xtrapolis 2.0 trains, we've implemented some temporary measures to keep our network running."
Tech Stuff
What's actually new in JavaScript (and what's coming next) This is a helpful introduction to what's new in ES2025 and what's coming in ES2026, and this writeup concludes with a handy prompt you can feed your AI tool to write ES2025 code.
I think I'm on my 5th different OAuth implementation throughout the years. They keep getting more and more convoluted. I don't really get the point of a specification when its so broad and complex that every service that implements it does it in a way that makes it incompatible with any other service.
Getting more out of OpenCode I've been using OpenCode for the last few weeks. In my experience it doesn't have all the capabilities of Claude Code, but it doesn't do a new release every day that breaks up a ton of features that worked before. There's something to say for simple and reliable implementations. And you can use OpenCode with different LLMs — Claude, ChatGPT, Gemini, and many more. I'm alternative between Deepseek V4 and GLM-5.1 both of which don't have excessive usage restrictions.
Vite+ I migrated my react-router apps to vite-plus. Part of it was easy switching, part took a bit longer (eg switching from Biome to OXC), but overall not super complicated and the performance boost is appreciated. With the help of AI I migrated 5 of my projects in half a day.
I Built My Dream Bug Tracking Software In A Week. 🎺
It is amazing to me that I’m able to use this fantastic bug tracking app that did not exist a few weeks ago. It’s mind-blowing. I am absolutely NOT a developer, but I have a good understanding of how development works on the web and am familiar with all the technologies needed to build an app like this. But I didn’t write any markup, any CSS, any JavaScript, nor did I do any design work (besides quickly throwing together that Mr. Bugbot head).
The question is not how fast someone can create software. The question is how long after creating the software will someone support it.
Open Design An open-source alternative to Claude Design:
Open Design (OD) is the open-source alternative. Same loop, same artifact-first mental model, none of the lock-in. We don't ship an agent — the strongest coding agents already live on your laptop. We wire them into a skill-driven design workflow that runs locally with pnpm tools-dev, can deploy the web layer to Vercel, and stays BYOK at every layer.
$0. $20. Then everything else. Last month I upgraded my Vercel free plan to the $20 Pro plan to get around usage limits and I just got my monthly statement for $41.33 because that's how Vercel works — they charge you whatever. There was no DoS attack, but if there was, you can't cap the costs. So I decided to switch to a different provider. I'll post about that soon. I just read this article which reaffirms that I should move away from the land of "Unlimited Charges":
Vercel's pricing page shows three columns. The invoice shows fourteen line items. This is what they don't put on the marketing page — the per-seat tax, the bandwidth overages, the DDoS-you-pay-for, and the hard cap that takes your free site offline without warning.
Quarkdown I'm not a big fan of making Markdown do things it was not originally designed for — hint: this is how XML turned into SOAP — but Markdown with some LaTeX extensions is not entirely bad idea (though, this spec tries to cover too much ground).
Datatype Font that turns simple text expressions into inline charts: type your sparkline as {l:10,40,25,70,50,90,35}, piechart it at {p:62}, etc. I love how easy it is to use this as readable text that gets turned into a nice chart.
Mike LLM for legal, an open-source alternative to Harvey and Legora. I think tech execs that need to review contracts, sign NDAs, etc would enjoy using this for its price and convenience. Pretty sure it's not for people in the legal profession.
BuhoBarX Another day another app that declutters the Mac menu bar. These are spreading because there's demand and shouldn't Apple Sherlock this category?
OrangeSlice Automate your GTM with AI: find customers, enrich data, run revenue workflows, etc. $20/month gets you about 400 emails and 40 calls, which I think is reasonable pricing.
DataCenter.fm Enjoy the background noise of the AI bubble without leaving the comfort of your office chair.
Eye for Design
Agent-First: Designing for MCP and CLI as the Primary Surface
Frontends aren’t going anywhere. Humans will still log in, look at dashboards, approve things, audit things, override things. That work stays.
What changes is which surface is primary. The default interaction with most software, over a long enough horizon, is going to be an agent acting on someone’s behalf. If you build for that surface first, the human one composes naturally on top of it. If you build for the human one first and treat the agent surface as an afterthought, you end up retrofitting forever.
Laws of UX A collection of best practices that designers should consider when building user interfaces: choice overload, cognitive bias, Fitts's law, Occam's Razor, et al. I find these laws so useful that I turned it into a skill, that way my AI tools can help me design better UX. Enjoy!
skillshare install https://github.com/cite-me-in/skills -s laws-of-ux
Side note: if you're asking your LLM to write a skill, after it does first draft, prompt it with something like "what are the goals you want to achieve? does it achieve these goals?" It will make suggestions, ask it to apply these corrections, and repeat the process 3~5 times. This process of revision-through-criticism works for blog posts, documentation, code, and more.
Peoples
AI Self-preferencing in Algorithmic Hiring: Empirical Evidence and Insights TL;DR For better chance of getting hired, ask your LLM to polish your CV:
To assess labor market impact, we simulate realistic hiring pipelines across 24 occupations. These simulations show that candidates using the same LLM as the evaluator are 23% to 60% more likely to be shortlisted than equally qualified applicants submitting human-written resumes, with the largest disadvantages observed in business-related fields such as sales and accounting.
Every "founder mode" essay is a man realising he's difficult to work for and deciding to publish an 8,000 word manifesto about why that actually makes him a Very Special Boy
We’re Training Students To Write Worse To Prove They’re Not Robots, And It’s Pushing Them To Use More AI (via Mr.Mark "The Sharpie King")
The irony of being forced to dumb down an essay about a story warning against the forced suppression of excellence was not lost on me. Or on my kid, who spent a frustrating afternoon removing words and testing sentences one at a time, trying to figure out what invisible tripwire the algorithm had set. The lesson the kid absorbed was clear: write less creatively, use simpler vocabulary, and don’t sound too good, because sounding good is now suspicious.
Business Side
Steve Blank AI and Teaching The times they be changing:
MVPs are No Longer an Indication of Technical Competence
Vibe coding has transformed MVPs to the equivalent of PowerPoint slides
Have y'all ever seen one of those horror movies where one of the characters ages like 500 years in ten seconds and crumbles to dust?
Me after reading this headline.
Machine Intelligence
AI outperforms doctors in Harvard trial of emergency triage diagnoses Not surprising given that it's a ton of information to process (medical knowledge) which favors tools that can do pattern matching at scale which is excatly what AI is good for. But the main takeaway should be that we're not paying nurses enough:
One experiment focused on 76 patients who arrived at the emergency room of a Boston hospital. An AI and a pair of human doctors were each given the same standard electronic health record to read – typically including vital sign data, demographic information and a few sentences from a nurse about why the patient was there. The AI identified the exact or very close diagnosis in 67% of cases, beating the human doctors, who were right only 50%-55% of the time.
My Adventures With ‘The AI That Actually Does Things’ Do you remember this book from your childhood?
Some clients, he said — people who just want to “soup up their experience with technology in general” and who are maybe feeling a little bit of AI FOMO — tend to focus on the idea of building a second brain. It strikes you that a lot of people probably either already have employees or wish they did. (The earliest LLM chatbots were especially appealing to people who enjoy ordering people around and being told they’re smart; the ability to assign tasks only makes the sensation stronger.) You joke that trying to isolate and construct elaborate workflows around your fairly simple routines makes you feel like a Richard Scarry character with a cartoon job in Busytown. Parker suggests that some of his clients are actually like Bananas Gorilla, the Scarry character who wears as many watches as he can fit on his arms.
Insecurity
Listening to cybersecurity people freak out over Mythos is so tiring. Like, bro, your local water treatment plant runs Windows XP, your mobile provider's hardware is older than you are, and the protocol that routes internet traffic is secured by everyone just agreeing that hijacking it would be uncool.
❌ Data breach
✔️ Surprise off-site backup
I just got given admin access to some Medicaid filing platform because I own the domain internaluser.com
Everything Else
18 silly finalists from the Comedy Wildlife People’s Choice Awards
First, I just want to say that I am offended by the assertion that 2000 was more than 25 years ago. That is a vicious lie.
Last night 3YO told a spoonful of rice that they should get excited because they were about to go see all their friends in her belly
After careful analysis, we believe the best option for remediation is to turn off the computers and go for a nice walk. Maybe call your mother.
Cursor Camp I got extremely productive, if by productive we mean attending Cursor Camp. It's a simple game that's very delightful, low pressure, and just fun.
some personal news: i recently put a power tool back where it was supposed to go and was able to find it with ease when i needed to use it today
You know if you search for "napping" in Indeed you get ZERO job results. It's almost like it's not even a desired skill.
I may be old but I still have cat like reflexes, I've knocked three different things off the counter already today.
Typibara Your work buddy you didn't know you need: lives at the bottom right of your screen and types along with you.
I love @tailscale. Whether I'm at a coffee shop or a doctor's office, I'm safely routing traffic back home, with access to my servers, my local LLMs, and even my cloud servers. The fact that I get to work there with a great team is even better.
Did you know that a cactus can be a source of water in an emergency?
If you encounter someone who won't share their water, repeatedly hit them with a cactus until they give it to you.
Follow me for more desert survival tips!
New Browser Plugin Adds Typos to Your AI-Generated Emails to Make Them Look Real It has come to this:
Now, in an effort to reintroduce some of the messiness of human writing — and hide our AI addiction — venture capitalist Ben Horwitz used Anthropic’s Claude AI to vibe code a browser plugin that does something that would have seemed preposterous just a few years ago: intentionally adding typos to emails.
Electrical current might be the key to a better cup of coffee
Their experiments confirmed that adding a single squirt of water to coffee beans before grinding can significantly reduce the static electric charge on the resulting grounds. This, in turn, reduces clumping during brewing, yielding less waste and the strong, consistent flow needed to produce a tasty cup of espresso. Good baristas already employ the water trick; it’s known as the Ross droplet technique. But this was the first time scientists had rigorously tested that well-known hack and measured the actual charge on different types of coffee.
Wesley Moore "Age verification"
South Africa withdraws AI policy due to fake AI-generated sources And now let's enjoy a moment of pure self-own:
JOHANNESBURG, April 27 (Reuters) - South Africa has withdrawn its first draft national AI policy after revelations that it contained fictitious sources in its reference list which appeared to have been AI-generated.
"The most plausible explanation is that AI-generated citations were included without proper verification. This should not have happened," Minister of Communications and Digital Technologies Solly Malatsi said.
My dad is a journalist. He wrote about a pilot project with self-driving busses recently. And something that none of the men I have talked to seem to realise is that it is fucking scary to think that you are in a bus alone at night without a driver and then someone starts harassing you. My male friend said "what is the bus driver gonna do?" but honestly, just the presence of another person is often enough. And a bus driver can call the police or stop the bus if need be, there are options.
This is 'Coppiplatz Bahnhof' Gouache on watercolour paper. I really love this one.
If you'd like a print, you can get one here
