Exporting Chrome’s reading list
I found that I had a few hundred saved links in my Chrome reading list, so I vibe-coded a quick way to export the links so that I could crawl each one, sort them and actually figure out which ones to read. I also use multiple Chrome profiles (personal, family, work, investments), so the script shows a summary of your profiles and allows you to choose which ones you want to export.
┌───────────────────────────────────┐
│ │
│ Export Chrome Reading List │
│ │
└───────────────────────────────────┘
↑/↓ navigate • enter select • esc quit
▸ Profile 1 (12 items)
Profile 5 (150 items)
Profile 8 (3 items)
✓ Exported 150 URLs to reading_list.csvFind it at: https://github.com/naveen/export-chrome-reading-list
In need of stories
I loved this recent post by @ashleymayer on how we need better stories about the future in tech – and small companies should be the ones to step up to tell them.
Just because capital is concentrated in a few of the biggest startups (nearly all AI companies) doesn’t mean they get to be the only ones to tell the big stories and use their larger megaphones. They keep telling stories about which model is the best, which one is growing the fastest, who has the most Github stars and so on.
There are all sorts of great insights in her post, but one in particular stood out the most to me:
Many of this technology wave’s most impressive companies have also made what I believe is a profound narrative error. They’ve cast themselves as the heroes in their own stories, and in doing so, risk becoming the villain in everyone else’s.
Historically, the best brands have made someone else the hero of their story. Apple was in service of the creative misfit, Nike celebrated the everyday athlete. When you build a story around your company as the hero, you risk turning your customers or users into NPCs. It signals an inherently transactional relationship, or worse, predatory (in the case of AI or robotics: we’ll replace you, just give us time).
The best brands make someone else the protagonist. Somewhere, tech (and, sometimes, those that write about tech) have lost that idea.
Additionally, the AI wave right now is perhaps in need of the same type of storytelling that the climate wants:
Telling this story requires a different way to tell a story […] As Wallace-Wells writes, we need an alternative: many problems we face now aren’t just one person’s problems where they go out into the world, selfishly solve it for themselves and come back home victorious. Most big problems are hard to define and hard to tell stories about. Global climate change, in particular, is known as a super wicked problem. We just may need some super wicked stories.
We all want to know what comes next: what happens to our jobs, what will we be doing, what does a new kind of information abundance mean, how does creativity, taste and the human side of things fit into it?
Terminal romantics (It feels like play)
There’s a specific feeling I remember from the early days of the internet — maybe 1993, 1994, somewhere in there. It was shortly after we moved to the US and bought our first computer. People were making things and trying things online just because: ASCII art. Chat bots. Personal homepages about, well, whatever, because you knew you just had to have a presence online, you knew you had to play in order to be a part of it all, to not get left behind. It was early enough that you got to try it all – BBSes, Gopher, WWW – so early that you didn’t know which of those methods to connect online was going to “win” (or, which would still be a cool, second place gathering spot). A lot of it was text-based and inside terminal interfaces.
It felt like play – a game.
I got that a little bit of that same feeling during the crypto years of 2020-2022. Everyone stuck inside during COVID, playing with money that didn’t feel quite real. (What’s the harm in trying stuff with house money?). Most of it seemed crazy (apes on a (blockchain) plane?) and some of it mattered (stablecoins). All of it had that same energy: people doing weird things because it was fun and the ceiling wasn’t visible yet.
The state of AI feels exactly like that for the past few months. Open a terminal, fire up claude or codex and start playing. Take cool ideas, half-baked concepts and try them out, just because. Text your openclaw agent anything and everything. You don’t necessarily know which approach or model or framework is the one that’s going to win, but you may as well play with them all. The cost to trying new ideas is low and so much fun to boot.
The only difference this time is the play is also the work.
The early internet was playful but the “useful” took probably the rest of the decade to arrive for everyone. Crypto was playful but for most people the useful arguably never really came. With AI, both are happening at the same time. We’re actually shipping ideas and features faster. Not a day goes by where a friends/parents group thread or team conversation isn’t talking about how to make the most of it. The game is producing real output.
By the way, given a lot of it is now happening in the terminal, I’d get your prompts to use Bubble Tea (or Gum) from the team at Charm. They make some really cool open-source tools for building beautiful terminal UIs. (* I am a small investor.)
West side space
Anyone I know have cool office space in Culver City or thereabouts? I want to spend one day a week on the west side and would be great to have a desk.
(Culver reminds me of Dogpatch; feels like the cool place for startups & tech right now)
In return: jam/hack sessions; help you with something you’ve got going on; welcome to hang with us in WeHo.
What I’m reading (Work)
Kalina on (endurance) durational art
Last week I went to see Tehching Hsieh’s retrospective at Dia Beacon. The show is called Lifeworks 1978-1999 and it’s up through 2027.
Have you ever heard of Tehching Hsieh? He is, in my opinion, one of the greatest durational artists of the 20th century.I had never heard of Tehching Hsieh when I started my “everyday” project. I was 20 years old and I just thought it would be interesting to take a picture of my face every day. That was basically the whole idea. It wasn’t influenced or inspired by anyone.
I would only learn about Hsieh later. First, I learned about the Time Clock Piece. From April 11, 1980 to April 11, 1981, he punched a time clock every hour, on the hour, and photographed himself each time. 8,760 punches in a year.
Karpathy on AI exposure by occupation
You are an expert analyst evaluating how exposed different occupations are to AI. You will be given a detailed description of an occupation from the Bureau of Labor Statistics.
Rate the occupation’s overall AI Exposure on a scale from 0 to 10.
AI Exposure measures: how much will AI reshape this occupation? Consider both direct effects (AI automating tasks currently done by humans) and indirect effects (AI making each worker so productive that fewer are needed).
A key signal is whether the job’s work product is fundamentally digital. If the job can be done entirely from a home office on a computer — writing, coding, analyzing, communicating — then AI exposure is inherently high (7+), because AI capabilities in digital domains are advancing rapidly. Even if today’s AI can’t handle every aspect of such a job, the trajectory is steep and the ceiling is very high. Conversely, jobs requiring physical presence, manual skill, or real-time human interaction in the physical world have a natural barrier to AI exposure.
(AI) Power to the people.
This is why retirees are lining up in Shenzhen. This is why people with no GitHub account are showing up at ClawCons. For the first time, they can feel AI’s intelligence, even if it is not very good. Yet. Not a demo. Not a keynote promise. Not big boys burning billion dollars a month. A thing that actually does things on their behalf. The gap between what you want done and what gets done has always required either your own time or someone else’s labor. OpenClaw makes that gap feel smaller. That feeling, even in its rough and half-broken form, is new.
It has been almost a month since I published How AI Goes To Work. “What OpenClaw shows is how AI will work in the background,” is what I wrote. “And that is what the ‘AI’ future looks like for normal people. Not a separate AI app. Intelligence woven into tools you already use. Doing work you used to do yourself. Or used to hire someone to do, done by software.”
MicCheck (Testing 1 2 1 2)
I wanted a quick system-wide menu item that would show my microphone’s current muted state and I wanted an easy way to change the state globally from this very same menu item. Sure, I can do this from a specific app (Zoom, Teams, Meet, …) but I didn’t want to go hunting from app-to-app looking for the mute button.
So: MicCheck. It is a tiny macOS menu bar app with one job: keep your microphone muted until you decide otherwise. It sits up in the menu bar showing ON AIR or OFF AIR. When something tries to unmute your mic without permission — a browser, a background app, whatever — MicCheck catches it at the system (CoreAudio) level and reverts it within milliseconds.
I built it with one feature that I couldn’t find elsewhere: a whitelist for allowed apps. For example, I use Superwhisper constantly for voice-to-text. The problem with a hard mute enforcer is that it would block Superwhisper too. So MicCheck has an allowed apps list — you add Superwhisper (or any other app you trust), and MicCheck steps aside when that app needs the mic. The moment your whitelisted app finishes recording and releases the input device, MicCheck re-mutes automatically. No button press. No forgetting to mute again. It just goes back to where it was.
The whitelist works at the audio session level, not just the mute property level. Most apps don’t touch the system mute flag — they open an audio stream and expect audio to flow. MicCheck watches for that too, so whitelisted apps get real audio while everything else gets silence.
It’s built with SwiftUI and CoreAudio, targets macOS 13+, and lives entirely in the menu bar — no Dock icon, no windows unless you open Preferences. Global hotkey (⌥⇧M by default, fully remappable), optional sounds and notifications, launch at login.
The source is on GitHub. Build it yourself or just grab the app download.
What I’m reading (AI reads)
Imagine describing AI to an ancient human –– “a superintelligent invisible being designed by the body of all of humanity’s recorded expression that helps us be productive, less lonely, and guides us through work and personal life.” –– almost any person in almost any civilization in the world would shrug and say “spirits, angels, devas, dybbuks, gods. Sure, no big deal.” We have, in fact, 100,000 years of robust and time-tested systems of organizing our society based on the belief that there are higher powers than ours. These higher powers move among us, determine how we all should act, and with whom we should be in communication — this is older than just about any principle we have together.
I think it is no coincidence that at the historical moment that humans progress themselves to the point of not breeding because it is inconvenient, that they invent a million virtual beings, a billion artificial minds, trillions of robots and a zillion working agents. Think of this as a handoff – a shift from one regime based on the biologically born to another based on the manufactured made. We are in transition from the world of the Born handing off to the world of the Made.
…
The purpose of handing the economy off to the synths is so that we can do the kinds of tasks that every human would wake up in the morning eager to do. There should not be any human doing a task they find a waste of their talent. If it is a job where productivity matters, a human should not be doing it. Productivity is for robots. Humans should be doing the jobs where inefficiency reigns – art, exploration, invention, innovation, small talk, adventure, companionship. All the productive chores should be handled by the billions of AIs we make.
The DeepSeek episode highlights another, arguably more revealing part of Nadella’s thinking: AI is rapidly commoditizing, and this is a good thing for Microsoft. While everyone in Davos was focused on AI consumption, Nadella was contemplating the history of coal production. One of his favorite economic theories is the Jevons paradox, which posits that as a resource becomes more accessible and its usage more efficient, consumption increases. This happened with coal during the 18th and 19th centuries and more recently with plane travel, when plummeting operational costs and airfares helped create frequent flyers, new flight destinations and booming sales for airlines. Nadella believes a similar phenomenon will play out with AI.
America Isn’t Ready For What AI Will Do to Jobs
Taken together, these statements are extraordinary: the owners of capital warning workers that the ice beneath them is about to crack—while continuing to stomp on it.
It’s as if we’re watching two versions of the same scene. In one, the ice holds, because it always has. In the other, a lot of people go under. The difference becomes clear only when the surface finally gives way—at which point the range of available options will have considerably narrowed.
One thing that should be learned from the bitter lesson is the great power of general purpose methods, of methods that continue to scale with increased computation even as the available computation becomes very great. The two methods that seem to scale arbitrarily in this way are search and learning.
The second general point to be learned from the bitter lesson is that the actual contents of minds are tremendously, irredeemably complex; we should stop trying to find simple ways to think about the contents of minds, such as simple ways to think about space, objects, multiple agents, or symmetries. All these are part of the arbitrary, intrinsically-complex, outside world. They are not what should be built in, as their complexity is endless; instead we should build in only the meta-methods that can find and capture this arbitrary complexity. Essential to these methods is that they can find good approximations, but the search for them should be by our methods, not by us. We want AI agents that can discover like we can, not which contain what we have discovered. Building in our discoveries only makes it harder to see how the discovering process can be done.
Blasting through X (Twitter) bookmarks
After years of accumulating tweets, I decided to tackle the monstrous backlog of 1,237 bookmarks on Twitter. (I’m almost kicking myself for not waiting just a few more days to hit the 1,337 marker.)
However, the web and mobile interfaces are painful for loads of content: slow to load, can’t group and categorize after the fact, can’t quickly open them in new tabs (“_blank”) and, worst of all, you can’t access these via API.
So, I used a simple Chrome extension to export the raw list of tweets. Then, with a little help from Claude, I wrote a Python script to do the heavy lifting: export the data, automatically sort and group each into distinct categories, and put them into a more readable format. The goal was to create a single-serving, local page that would let me finally blast through my reading list in a weekend.
My categories and counts reveal a lot about what captured my attention over the years, and honestly, the distribution probably still holds true today:
Investing & Finance (387), AI & Machine Learning (264), Other (103), Startups & Entrepreneurship (82), Productivity & Tools (77), Long Reads & Essays (70), Career & Professional Growth (56), Media & Entertainment (55), Design & Creativity (54), Tech Industry & News (44), Society & Culture (21), Health & Longevity (13), Images & Visual Content (8), Humor & Fun (3)
Reviewing six or seven years of bookmarks was a fascinating trip down memory lane. It was interesting to see which posts and articles have genuinely stood the test of time. And, which haven’t: oddly, a lot of crypto content. I bookmarked many crypto tweets during the intense 2020–2022 period because it was all moving so quickly that it was impossible to keep up. In a way, the current flow of AI (Code! Cowork! OpenClaw! Tunnel straight into your Mac mini at home!) updates feels similar. A constant stream of new information you probably want to save-for-later. Except this time, I’m not: I’m reading them because it’s so much fun to be building using these tools.
If you’re ready to tackle your own archive, here are the two tools you’ll need:
- Use the X Bookmarks Exporter Chrome extension to export a CSV of all your bookmarks.
- Feed the output to the x-bookmarks-viewer python script to process the data and design a single-serving page locally. You can fork it, drop in your own CSV export, run
'python process_bookmarks.py`, and have your own organized bookmarks viewer.
All Watched Over by Machines of Loving Grace
From 1967:
I like to think (and
the sooner the better!)
of a cybernetic meadow
where mammals and computers
live together in mutually
programming harmony
like pure water
touching clear sky.I like to think
(right now, please!)
of a cybernetic forest
filled with pines and electronics
where deer stroll peacefully
past computers
as if they were flowers
with spinning blossoms.I like to think
(it has to be!)
of a cybernetic ecology
where we are free of our labors
and joined back to nature,
returned to our mammal
brothers and sisters,
and all watched over
by machines of loving grace.