Labnotes

Published

Weekend Reading — So close to the end of the world

Weekend Reading — So close to the end of the world

Laura Manach


Tech Stuff

Convert & Compress When you need to process a batchload of files: convert them to a different format, limit their width/height, compress with specific quality loss, etc.

(* To make sure my webapp is performing well, I run a script that goes through all files in the directory, down-sizing images so not wider than 1024px, compressing with 90% quality, and saving as JPEG )

Helping people write code again

If you have significant previous coding experience - even if it's a few years stale - you can drive these things really effectively. Especially if you have management experience, quite a lot of which transfers to "managing" coding agents - communicate clearly, set achievable goals, provide all relevant context.

Web development is fun again

They’re far from perfect, but claude and codex gave me the leverage I desperately needed. They’ve brought me back to levels of productivity I haven’t felt in years. I feel like I can manage the entire stack again - with confidence.

I can go from idea to execution in days.

Suddenly, the complexity of each domain matters a lot less.

The November 2025 inflection point

It genuinely feels to me like GPT-5.2 and Opus 4.5 in November represent an inflection point - one of those moments where the models get incrementally better in a way that tips across an invisible capability line where suddenly a whole bunch of much harder coding problems open up.

Directional CSS with scroll-state(scrolled) Interesting way to dynamically change website style based on scroll behavior, position, direction, and more.

Opus 4.5 is going to change everything 🤔

What you don’t need: variable names, formatting, comments meant for humans, or patterns designed to spare your brain.

What you do need: simple entry points, explicit code with fewer abstractions, minimal coupling, and linear control flow.

Why AI is pushing developers toward typed languages On the other hand (there's no one right answer):

Type systems don’t replace dynamic languages. But, they have become a common safety feature for developers working with and alongside AI coding tools for a reason. As we see AI-assisted development and agent development increase in popularity, we can expect type systems to become even more central to how we build and ship reliable software.

Repomix I'm trying out Repomix to see whether it makes Claude Code smarter (as in faster to write correct code).

b0df01fbd56f7ba2.png

MCP is a fad I think MCP will prevail, but not every technology that prevails is optimal:

This code was a hassle to write, prior to the advent of coding agents. But these small utility scripts are the precise thing that coding agents excel most at! A technical user of MCP tools will be hard-pressed to find a tool an agent could not one-shot in the programming language they are most comfortable in.

6 incredibly hyped software trends that failed to deliver

Others watched blockchain get replaced by simpler tech that just worked. “One supply chain project I saw was shelved after a year and replaced with a simple setup: [Apache] Kafka, signed records, and [Amazon] S3 immutability,” says Srikara Rao, CTO of cloud and cyber security services at R Systems, a global digital solutions provider. “It lacked the blockchain buzzword, but worked reliably and scaled.

In hindsight, what were we thinking?

Stop Doom Scrolling, Start Doom Coding: Build via the terminal from your phone I don't recommend doom coding but I love that in 2026 this is something you can choose to do on a small budget.


Eye for Design

It’s hard to justify Tahoe icons A wonderul writeup about the horrible design mistake that is macOS 26: (via Thomas Steiner)

Fast forward to 2025. Apple releases macOS Tahoe. Main attraction? Adding unpleasant, distracting, illegible, messy, cluttered, confusing, frustrating icons (their words, not mine!) to every menu item.


Peoples

Your team has a context window How to design team scope while keeping the team healthy and efficient:

If a team constantly feels overwhelmed, reactive, or trapped in meetings, it’s rarely because the people are underperforming. It’s because the team is attempting to operate with a context window that exceeds the team’s cognitive design.

The fix is not more process, or more meetings, or more dashboards. It’s a deliberate shrinking and shaping of the context window so the team can think again.

How I hire engineers 👍

In both cases, the questions the candidate asks are almost more important than the questions the panel asks. They, once again, speak to how the candidate thinks; what they’re curious about; what’s important to them.


Business Side

The quarter we treated SEO like bookkeeping and did just enough to stop it hurting us later

The lesson for other founders wasn’t that SEO must be your primary channel from day one, but that doing nothing at all is a compounding liability. Treating it like bookkeeping consistent, minimal, and boring, prevented it from becoming a massive catch-up project when we finally needed inbound to matter.

Get Rid of Your Boss. Share a Story Instead.

Most organizations don’t run on org charts. They run on stories. If you want to replace hierarchy with real self-management, you must first replace the old story with a new one.


Machine Intelligence

AI isn’t “just predicting the next word” anymore

The latest models are more “path-finder” than “next-word predictor.” At any given moment, these AIs aren’t really trying to predict what word comes next at all; they are selecting the next step to solve a problem, and then executing that problem-solving step; they are pursuing a strong answer in a wide variety of ways, rather than outputting a response directly. (More generally, this type of path-finding AI is called a “reasoning model.”)

Why AI Companions Feel Addictive (And How To Spot It Early) (via Michael Martinez)

These signs aren’t meant to shame anyone — they’re simply patterns I’ve noticed in clinical practice that, when caught early, are much easier to shift. If some of these feel familiar, please be kind to yourself. Reaching for comfort in a difficult moment is deeply human.

fancysandwiches

One of the ways I'm dealing with AI slop at work is that when I'm giving feedback on the work I'm making sure to never assign the responsibility of the bad code to the AI. I'm directly saying that "this change that YOU made needs to be corrected". I'm always assigning the output of the AI to the person who put me in the position of reviewing the work. It is their responsibility to read the code that they're trying to review, they are responsible for 100% of the code, so they also get 100% of the blame when it's bad. If a change is confusing or nonsensical I'll ask "why did YOU make this change?". I'll never ask why an AI made a change, that we cannot know. All we can know is why someone thought it was acceptable to ship garbage, and we can assign them the responsibility for the garbage that they're willing to ship

LLM Problems Observed in Humans Humans don’t know when to stop generating, have a small context window, too narrow a training set, and other failure modes typically associated with LLMs:

This has always been an issue in conversations: you ask a seemingly small and limited question, and in return have to listen to what seems like hours of incoherent rambling. Despite exhausting their knowledge of the topic, people will keep on talking about stuff you have no interest in. I find myself searching for the “stop generating” button, only to remember that all I can do is drop hints, or rudely walk away.

Vertical Rickroll AI rewrites your text into an acrostic masterpiece. The first word (or letter) of each line spells out a hidden message—like the classic Rickroll, movie quotes, or viral trends.


Insecurity

Simo Kohonen (via Ian Campbell) 😮

Ragebait as a phishing tactic.. a threat actor pretending to be SendGrid is sending out phishing emails with the note: '..We will be adding a "Support ICE" donation button to the footer of every email sent through our platform'.

Feels like the de facto way for phishing people was always trying to lure people with something pleasant, like a list of Christmas bonuses for the year - but psychology has known this effect for a long time, i.e. losing $100 hurts more than finding $100 feels good, or in this case, appearing to support ICE hurts more than knowing the bonuses of the IT department feels good.

Not Just Bikes

I remember when the DMCA was first being proposed and I was trying to tell anyone and everyone that anti-circumvention would be disastrous. I must admit that I had no idea how bad it would be.

I knew that media companies would use it to enforce DRM, but I had no idea that John Deere would use it to brick tractors or that Medtronic would use it to stop hospitals from repairing ventilators.

Wendy Nather "You gotta wonder whether these scam farms collect and share the best reactions to these opening lines"


Everything Else

batkaren "Sadly, we must live in the apocalypse we’re dealt, never the one we desire."

Molly White

i love the beginning of the year because everyone starts blogging. and if you (yes you) were thinking about starting, this is your sign

Jerry

There is some weird time dilation BS going on. I’ve been on the elliptical machine for at least 2 hours now but the machine is only saying 10 minutes 😱

Melancholic Mediocrity

So if you think about it, a black hole is really just a gigantic Everything Bagel

Obsolete Sony "I miss Weird Sony"

Staff Chief of Joints

"Naturally flavored with other natural flavors" are words I just read on food packaging and I feel like there's a wink at the end of that.

Alessandra Sierra

“Speedy Meetings” in Calendar is Google’s greatest contribution to productivity

(Speedy Meetings: automatically set events to end 5 or 10 minutes before the next time block, so a “1-hour” meeting is actually 50 minutes)

mhoye

Daughter just came home to tell me the teachers at school tried to confiscate her vape but they couldn't because it's actually a kazoo.

"I made a kazoo noise at them and I could see them dying on the inside so I made a sad kazoo noise instead. I don't think it helped."

The grin on her face.

Stay on Track With Your Health If you need an Rx reminder that gives you more flexibility in changing schedules and an historical view.

Li Chen

I hate when I'm walking and there's a bird ahead and it tries to get away but it just flies further in the direction I'm going so it has to keep flying and landing and flying and landing and now the bird thinks I'm a weird bird stalker and will probably tell all the other birds😔

Bose is open-sourcing its old smart speakers instead of bricking them 👏

Now that the smart speakers’ API is being open-sourced, users can also create their own compatible SoundTouch tools to help fill in any gaps left by the lack of cloud services. While it’s still disappointing that the speakers are losing official support, Bose’s approach at least lets people continue using their speakers, rather than bricking otherwise functional devices.

Closing Central Madrid To Cars Resulted In 9.5% Boost To Retail Spending, Finds Bank Analysis (via Anna Anthro)

Things You Realize When You Run Out of Fucks to Give re: fuck depletion

You’re not sitting down with a spreadsheet. You’re not making pros and cons lists.

But you start noticing who you genuinely want to see and who you’re seeing out of obligation. Who lights you up and who drains you. Who accepts you and who needs you to be smaller.

And then, this is the wild part, you just start acting on that information.

The 'save for later' paradox: why we hoard digital content we never read My always-growing-never-shrinking Instapaper feed in a nutshell:

Digital clutter pretty much resembles physical clutter. Just as some people hoard physical items "just in case", others tend to hoard digital content for a future that never comes. Why do people continue to save content when they know that they probably will not get around to consume it? Do we all have some hoarder tendency?

John

I wish we had spent the last 26 years teaching people that the reason the 2000 bug didn't destroy a significant amount of our infrastructure is because we caught it and spent thousands of hours fixing it BEFORE the year 2000

Because within that little perplexion - people thinking the problem was a hoax because it was fixed before it destroyed shit - is an encapsulation of the current era of Western politics, including COVID mitigation, lesser evil politics, fascism, and crime rate hyperbole

How To Stop Being Lazy And Get More Done: 6 Secrets To Motivation

A specific goal is rude. It’s a number, a deadline, and it has the gall to require you to notice whether you did it. It doesn’t care that you meant well. It cares that you ran two miles, wrote the document, made the appointment, or didn’t.

A clear goal gives the world (and more importantly, gives you) a way to find out if you’re full of it. That’s why it works.

Stef Walter

When we design for disabilities, we make things better for everyone. This is called the Curb-Cut Effect. The term was coined by disability students and activists in the 70s, who added curb cuts to the Berkeley sidewalks to make access easier for those in wheelchairs. They discovered those also helped people with strollers, using trolleys for deliveries, etc.

🔥 Looking for more? Subscribe to Weekend Reading.

Or grab the RSS feed