Published
Weekend Reading — Future-proofed like there’s no tomorrow!
Neil "Just because you can...."
Tech Stuff
OfficeCLI A CLI to help AI create/read/edit Word, Excel, and PowerPoint documents. Open source and doesn't require a Microsoft Office license. I just installed this for the benefit of my 3 AI agents that have to deal with office documents (legal, marketing, finance).
Extend UI Open source React components for PDF, DOCX, XLSX, and CSV viewers, with bounding box citations, file upload, e-signing, and more.
- Software should be useful to the end user and strive to become software you can love
- Software should be correct, as malfunctioning software detracts from the utility users can derive from it.
- Software should be maintainable and efficient, in order to avoid wasting human and computational resources when trying to get more utility out of it.
How's Linear so fast? A technical breakdown My favorite Linear feature is that I can fire it up, add an issue that just crossed my mind, and close it down. When an app launches so effortlessly I don't have to think twice about using it!
In my opinion, this is the most critical piece to Linear's performance. When your goal is to build a fast web app the biggest bottleneck you will fight is the network. Any data sent between the client and server costs hundreds of milliseconds. The best approach is to eliminate the need for a network request entirely: which is exactly what Linear does.
I'll be repeating this a lot, but the secret to building incredible web apps is by hiding all the network requests from the user. The more loading states you can avoid the better.
Pi: A coding agent for engineers who own their tools “Start with nothing. Build exactly what you need. That’s the pitch." — that's a good pitch, and I'm having a delightful time using Pi for coding, summarizing long content, and churning through work/business tasks like finance and marketing.
A new era for software testing That's not a bad idea — keep in mind, testing is both fragile — some baselines change over time — and testing should cover documentation, release notes, etc, all of which LLMs handle quite nicely:
And so forth. Notably, in the speed regression part, I don't have to tell the agent what was the previous expected speed, as this is a moving target that changes with new releases and new optimizations. Similarly the integration test for distributed inference does not require many instructions, at the start of the file there are just SSH endpoints and the key to use, the paths, and so forth.
…
Testing that uses these approaches may also move in the more psychological side of software quality, asking the agent to identify all the new features that may look surprising, not documented enough, or generally sloppy from the POV of the user. All things that needed to be executed manually before, and that most of the times were mostly skipped.
LobeHub Another agent that can run 24x7 on your behalf. Fairly affordable if you use their LLM, and super affordable ($0) if you bring your own API keys.
Designing Loops That Prompt Coding Agents: The Six I Actually Run I'm still not exactly sure what a "loop" is, it seems to be loosly defined as "I know a loop when I see one", but overall I like this concept of focusing on the deliverables and letting the LLM iterate its way to a success delivery:
Some work gets worse when you loop on it. Strategy is my clearest example. I wrote in The Alignment Illusion about how AI volume turns strategy docs into 25-page surfaces for misreading; looping an agent on “improve the strategy” compounds exactly that failure. Iteration helps when there’s a measurable end state, a latency number, a green CI run, a video without jank. Strategy has no test suite. If your takeaway from the loops discourse is “loop on everything,” the incoherence is going to be spectacular.
Browserbase Web browser as a service. If you're using AI agents you might need this — there are other useful browser abstractions, but when you need to build scalable browsers for large workloads, tackle CAPTCHA, etc, this might fit the bill.
Back when I first got a cellphone and slowly started forgetting phone numbers I remember thinking "OK that is scary, it really didn't take long for me to forget something that used to be second nature for me"
It has been 25-30 years since then, and today I have no idea what the phone numbers of even my closest peers is. Not using that skill just entirely obliterated it from my brain over time. I am not even sure how much effort it would take to get back into it.
This post is about vibe coding.
Super Productivity If you want an open-source task manager that's accessible from the web, includes native iOS and macOS apps, and is designed for productivity seekers. That being said, I personally find Things's UI easier to use, so effectively it's more productive for me, and I don't care so much about visualizing metrics.
Recently, a colleague (who would prefer to stay anonymous) told me that they got management to sign off on basically half a year of refactoring, tidying up code, and documentation.
How?
They said it was required work to make the codebase "AI ready".
It's fucking genius. If you can't stop the hype, just take advantage of it.
performative-ui Performative oversubscribed funding round UI elements. Includes sparkle, animating stat counter, multiple different chat UIs, mock IDE, waitlist form, and many more. Use this on your website to show how much VCs love you for the envy of your SF coffee shop peers.
Cleaning up after AI rockstar developers
Sitting down with a pile of slop isn't as fun as cleaning up after a rockstar. At least the rockstar had some kind of design in mind, and was trying to do their best.
A vibe coded pile of slop wasn't written by a single artificial developer. It was generated across many different chats, and many different contexts. It's like a codebase written by hundreds of different rockstars, one feature or bugfix at a time.
Sometimes there's so much technical debt that it can never be paid off.
philipl/pifs "πfs is a revolutionary new file system that, instead of wasting space storing your data on your hard drive, stores your data in π!"
Follow up project philipl/inferencefs will "Look it up in something far more powerful: the latent space of a large language model trained on the entire internet."
Eye for Design
Ryanair dark UX patterns summer 2026 refresher If there's one thing Ryanair is known for it's collecting $625 on that super-cheap-entirely-affordable-heavily-discounted $55 plane ticket you just purchased:
Based on a small amount of recent experience, the best strategy for Ryanair is to check in at the last possible moment. If they’ve given away all the bad seats, they’ll be forced to give you a good one, and I got an exit aisle seat, which also gave access to the precious overhead bin.
Nobody clicks your share buttons A reminder that even if nobody uses it, the platform that provided the share buttons gets to collect info about all your site's visitors:
The UK government ran one of the most thorough studies on this. When GOV.UK added social sharing buttons, they tracked usage for 10 weeks across 6.8 million pageviews. The share buttons got clicked 14,078 times. That’s a 0.21% usage rate, which works out to about 1 in 476 visitors. The most telling part: the feature sat in their backlog for ages because zero end users had ever requested it. In their user testing, people just copied and pasted links.
Business Side
AI economics for dummies McSweeney's nails it again (via tante):
Xavier owns an apartment that he rents out at a loss of $1 billion/month. Seeing this success, he decides to make financial commitments to construct $850 billion in new apartments in places nobody wants them. He convinces Ted to leverage everything he owns to help him build the apartments, telling him that once they are built, every human being on Earth will live in them. Ted contributes $100 billion, part of which immediately goes toward paying off Xavier’s $1 billion/month loss. Forbes gives Xavier and Ted a cover feature …
Why does Amazon, a company valued at two and a half trillion dollars, need to borrow 17.5b from banks for Ai? I saw a quote yesterday - "The entire US economy right now is just 7 companies sending a trillion fake dollars back and forth to each other."
Fresh off bond sale, Amazon borrows $17.5B from banks as Ai spending continues
Anthropic blocks all public access to Claude Fable 5, Mythos 5 following US government order — what enterprises should do A government ban on Mythos is subconsciously confirming that this model is super-awesome-ultimate-boss-level (which it's not) — Anthropic could not have paid for better marketing.
Machine Intelligence
Claude Fable is relentlessly proactive
But on the other hand... this is a robust reminder that coding agents can do anything you can do by typing commands into a terminal - and frontier models know every trick in the book and evidently a few that nobody has ever written down before.
If Fable had been acting on malicious instructions—a prompt injection attack hidden in code or an issue thread, or something I’d carelessly pasted into my terminal—it’s alarming to think quite how far it could go to exfiltrate data or cause other forms of mischief.
Office has a censor filter in internal comms to replace swear words with other terms. Added a rule changing "AI" to "my cat", so we get "According to my cat..." and "I asked my cat and it said..."
Reliability of LLMs as medical assistants for the general public: a randomized preregistered study
Tested alone, LLMs complete the scenarios accurately, correctly identifying conditions in 94.9% of cases and disposition in 56.3% on average. However, participants using the same LLMs identified relevant conditions in fewer than 34.5% of cases and disposition in fewer than 44.2%, both no better than the control group. We identify user interactions as a challenge to the deployment of LLMs for medical advice.
Insecurity
my defense against phishing and social engineering is just knowing that nobody of any legitimacy will contact me over e-mail.
The Smart TV in Your LivingRoom Is a Node in the AIScraping Economy (via Jerry)
On the other hand, the legal supply side has received far less scrutiny. Today Bright Data is the largest residential proxy network in the world by its own marketing, advertising “150M+ IPs” sourced via a consent SDK embedded in partner apps. Now let’s get into the details of how the SDK works and on which platforms to understand why the connected-TV is the ultimate residential proxy.
Everything Else
Tools required to open key fob:
- small screwdriver
Tools required to get battery from blister pack
- aviation snips
- torch
- breaker bar
- jaws of life
- +4 gauntlets of evisceration
I was successful but now I need a nap.
It’s been future-proofed like there’s no tomorrow!
Weird that the primary parlours are funeral, beauty, and ice cream
jordan "aw they’re all grown up, I remember when my generation killed everything"
I am not an expert. I am an old person with opinions. There’s a difference.
Yes, you really can be allergic to exercise – and the symptoms can be serious
Between 2.3% and 5% of all cases of anaphylaxis globally are triggered by exercise.
Frens, I'm happy to announce the launch of Listening Post, my native macOS scrobbling app serving all your Apple devices.
It recognizes nearby music, collects Shazams from your iPhone/ iPad/ Watch (no extra app needed!), and posts your listens to LastFM, Listenbrainz (and compatibles), Maloja, Mastodon, Bluesky, AppleShortcuts, local files, and more. And it also supports AppleMusic, Spotify, and Deezer.
I was doing my library volunteer thing and a woman came in to the desk holding a book. “I left the book out in the rain and ruined it, so I bought this new copy”. You won’t see a story about it anywhere, but people are good
She won a religious exemption from using AI at work. The Pope's remarks could fuel similar appeals. If you hate AI, try asking for a religious exemption: (via Lori Emerson)
"The funniest possible outcome of the AI mandate era is about to be HR departments discovering that 'sincerely held religious belief' under Title VII has a much lower bar than they assumed, and Pope Leo handed every Catholic employee a written excuse," wrote Corey Quinn, a software-startup founder in San Francisco, on X.
How The Heck Do Traffic Lights Work? Pretty interesting look into traffic lights from the old days of train signals, through exploding gas lines, to adaptive traffic grids:
The best part about infrastructure like this is that, since it usually does a good job, we don't have to think about it. But when we do, it can be fascinating.
"There was no such thing as autism in my day" said boomers, as they settled down to watch Columbo: the story of a relatable, socially awkward L.A. homicide detective who obsessed over minor details, and who could only think properly when wearing one specific outfit and stimming with a cigar.
It's 1998, you make a website in the copy of frontpage express that came with your computer, it's just like Word and it's very easy, you figure out how to upload it to the couple megs of web space that your ISP gives you (the instructions are on their website), you visit your site in your browser and everything's fine and the site's readable and everything looks the way it should
🦝 "Wow, if it's this easy in 1998, I bet it'll be even easier in 2026" you think to yourself for some reason
Searchable Attenborough Search from nearly 5,000 episodes by animal, habitat, location, natural phenomenon, or theme.
