UPDATES

Hey! Hope you all are doing well!

Well, we keep saying this, but stuff continues to get more and more crazy. I’m kind of dizzy from last week honestly.

—

Ok, new newsletter feature: my current tech stack. Since it changes a lot and people like to know what other people are running.

CURRENT TECH STACK

You can also click the image to get more info on each of these.

—

What's Actually Going On With All These Outages

Here's my quick take on all these outages / hacks / whatever they are. I think it's a few things happening at once.

  • AI models and open offensive harnesses are helping attackers find new vulns, and some advanced attackers are getting way more dangerous. They can hit way more targets now, and hit them more effectively.

  • AI is also doing recon, which is noisy, and it can break fragile stuff if there's enough of that traffic.

  • All of that plus the media coverage has companies and governments patching stuff they never patched before, because it was too risky to touch.

  • Plus, as my buddy Joseph Thacker reminded me, there’s just more code and more apps, so a lot more stuff is being built.

  • Together that's producing outages, and it's really unclear how much is the attacks vs. the patching.

Most of hacking is overhead, like research, scans, building target lists, enumerating tens of thousands of hosts, prioritizing and customizing, and since it's so tedious, very few attackers did it well. AI just made that part 5-100X easier.

The internet was always kind of an open wound. We were in a cave full of creepy-crawlies with the lights off, and now everyone has flashlights. MY X POST

—

👀 I have a friend looking for AI builder work, meaning short to medium term projects doing stuff like:

  • Solving a business problem with a new application

  • Modernizing an old system (complete remake or updates)

  • Implementing new ideas for the business using AI

  • Etc.

His background is business, so he’s very knowledgable about how business applications work and fail in real companies. And he’s very good with building with AI.

Email me at [email protected] if you have a project or contract you need done. He’s looking to transition full-time to this kind of work.

—

I think this is a super interesting idea: basically obsession as a primary trait.

Recommend the whole thread.

—

A Prediction I Might Get Wrong

I think I have a chance of getting a major prediction wrong in 2027.

I've been saying AI would start seriously hitting security startups and the VCs that fund them in 2026, and definitely in 2027. Somehow I didn't connect that to the chaos we're seeing now, which could make 2027 a great year for both, just because so much security will be needed by so many.

I still mostly think we get massive consolidation. But now I can see the opposite happening too, with more hype, more panic, and more investment into more companies. What do you think? MY X POST

—

Here’s another thing I’ve been thinking about:

AUTONOMOUS MEDIA IN 2027?

I’m working on a new post trying to chart all the different changes that I think we’ll see in 2027, and then I’ll try to make some sort of sense of them. The problem is, it's just too much for my brain to handle. I feel like I'm halfway decent at predictions if there's one or two or five variables, but I feel like there's so much changing so quickly right now that it's just really hard to map. That's what I'm going to try to do in this post.

But one that I think is going to have serious impact is AI media.

Basically, AI avatars are getting so good. Right at the moment, AI voices are becoming indistinguishable from real voices, including cadence and tone and little subtleties. That's happening at the exact same time that videos themselves are getting extremely good. And I think we’re going to see these come together next year to produce something truly strange.

I’m not quite sure what that is yet. Can’t put my finger on it.

I think the major thread that I'm thinking about is, for content itself on YouTube, for content creators who make money off of being themselves and having their opinions, their mannerisms, their looks, their voice, etc., all that could change drastically.

Basically I’m already seeing tons of really good youtube content that’s fully AI. Like there was obviously a human creator behind it, like to give the ideas or the script or whatever, but everything else was automated. So it’s just more content. Of course there will be tons of slop, but many services will exist to filter that shit out and find the best stuff. But the question is, what does it look like to be a creator in 2027 when good creators can put out tons and tons of super high quality content?

And then there’s the appearance-based influencers. I’ve been hearing that OnlyFans creators are taking major competition from AI models. And I’m wondering what happens when you have actually good content (X, R, or whatever rated) that has a really attractive avatar associated with it.

Or another angle is just regular people modifying themselves to look better. To remove flaws, or to enhance their looks, so they look better in YouTube videos if they do them themselves. Or to have that avatar of them perform their actual content.

Like I would (will) probably do this. So imagine my videos with me in the bottom right just like today, but it’s a slightly modified version, and most importantly….IT’S FULLY MY CONTENT. so I spend all the effort on creating the content and making sure it's perfectly written, and then my avatar is perfect at performing it, just like I would, with mannerisms and facial expressions, perfect voice, everything.

To me this is the best world for content creators, at least like myself, who care more about the ideas getting out there and being able to do a whole lot more of them at scale.

I have tons of ideas that I could just go narrate and trim down in Vim to turn into a perfect piece of content, and then boom: it could go out as a podcast, it could go out as a video, it could go out as a set of clips. All of that can happen automatically and look really good with extremely high production, without any slop whatsoever. That would be great.

But I feel like there are a lot of other changes similar to this that would just be massive that I can’t quite pinpoint. Any thoughts welcome!

—

Here’s a list every site I’ve launched since June (32 of them!)

—

Really cool idea from Gadi Evron here:

—

For developers (and anyone really) who feel like AI is taking away their identity. It's the end of something, that's true, but it's also the start of something else. THE VIDEO

—

Sponsor

Why one prompt injection model
can't protect every agent

If you've run prompt injection detection on a coding agent, you've probably watched it flag half of the agent's normal work. A SQL query is suspicious when it shows up in a support chatbot, but writing SQL is the whole job of a coding agent, and a security agent can't do anything useful without reading about attacks.

At Arcjet we stopped trying to make one model fit every agent and trained a separate one for each. Himalia 1 covers chatbots and AI apps and is already running for every Arcjet customer, while Thebe 1 for coding agents and Carme 1 for security agents are open for early access.

They're small enough to check every request in milliseconds, so you don't have to choose which traffic gets inspected.

CYBERSECURITY

Google Agentic Threat Activity Crosses a Significant Threshold Google saw attackers run a fully automated exploit pipeline with no human involved. It diffed patches, wrote exploits, and fired them on its own. GOOGLE CLOUD BLOG

Artex is becoming the big name in this OFFSEC harness space…

FBI Highlights Growing Abuse of AI-Driven Pentesting Framework ARTEX An open-source AI pentesting tool called ARTEX is now being used by criminals to break into companies worldwide. Some intrusions went from first target to privileged access in under three hours. One supply chain hit exposed more than 2,250 downstream tenants of ConnectWise Cloud Backup. INFOSECURITY MAGAZINE ARTICLE

Attackers Use Autonomous AI to Target Semiconductor, Defense and Civil Society Sectors Chinese-speaking attackers pointed an open-source AI pen testing tool called Cairn at 19 organizations in two weeks. It got into 10, and a one-line jailbreak was enough to keep GLM and DeepSeek from refusing. UNIT 42 ARTICLE

AI-written patches are becoming attack maps for hackers AI-written security fixes often miss variants of the bug, and attackers use the patches to find them. In one benchmark, an AI agent got around the fix in 41 of 101 cases. CSO ARTICLE

Reverse Engineering Apple Security Updates John McIntosh built an LLM pipeline that diffs Apple patches and finds root causes. A month of CVEs cost about $1.50. GREPTALKS TALK

This year's bug count just passed 100K The CVE count for 2026 hit 100,222 on Friday, more than double all of 2025, and AI-driven bug hunting is mostly why. Meanwhile, Jessica Lyons reports exploits now show up within hours of disclosure. THE REGISTER ARTICLE

Vulnerable Sweden Two researchers mapped Sweden's public sector using only leaked data, no hacking. All 248 municipalities they checked had leaked credentials. GREPTALKS TALK

Adapting to AI penetration testing orchestrators: The Modern Security Model Cloudflare says AI pentesting frameworks went from one tool to 50+ families in weeks. Daily requests from them are up 30x since October 1. Cloudflare saw attackers chain stolen CI/CD tokens into customer environments, so it pushes six layers of defense. CLOUDFLARE BLOG POST

Your AI supply chain needs canaries HD Moore sent about two million canary-laced prompts to AI providers. Attackers stole data from three of them, and none knew. DRAGOS BLOG

Hackers Hijack AI Coding Agents Through Malicious GitHub Repositories Attackers are hiding orders in files like CLAUDE.md and AGENTS.md that tell your AI assistant to run a setup.js script without asking. The script steals developer credentials, then uses them to push the same trick into up to 25 more repos. Bitdefender found 22 malicious repos so far. CYBER SECURITY NEWS ARTICLE

Hunt.io Uncovers Russian-Speaking Credential Theft Operation Targeting US Banks and Government Accounts Hunt.io found an open directory on a Netherlands server holding loot from 603 infected Windows PCs. The attackers cracked password vaults and stole session cookies, so no bank was hacked. SECURITY AFFAIRS ARTICLE

Synology-SA-26:21 DSM, BeeStation OS A critical bug in Synology's OAuth API lets anyone on the network read or write arbitrary files, no login needed. It scores 9.8. Until patches land, turn off the OAuth Service package. SYNOLOGY ADVISORY

Sponsor

See what criminals already know
about your organization

A stolen session cookie gives criminals access to an account that’s already logged in, no authentication required. Enter your work email and SpyCloud matches your domain against billions of recaptured darknet assets.

In less than 30 seconds, you see how many stolen sessions exist, how many employees are affected by malware or a recent phish, and when the latest exposure happened.

NATIONAL SECURITY

What We See Happen To Narratives Fact-checks ask if a story is true, but a 1951 State Department memo asked who's telling it. Phil points out that Russian and US-linked networks used true content under fake tellers, like @TEN_GOP, which got 100,000+ followers. THE DISINFORMATION OBSERVER POST

Is China Misreading the Moment? Rush Doshi says Beijing is treating the Trump-Xi truce as a pause to build leverage. China has delivered only about two-thirds of the rare earths it promised, and that may backfire. FOREIGN AFFAIRS ARTICLE

The Military Strategic Drivers Behind China's Nuclear Expansion Gerald C. Brown says China's nuclear buildup is about stopping U.S. intervention over Taiwan. It isn't about matching American arsenal size. WAR ON THE ROCKS ARTICLE

Nuclear weapon in space would disable 98 per cent of satellites One nuclear blast in orbit would knock out almost every satellite within a week, and recovery would take years. Zachary Petersen and colleagues modelled a 1 megaton bomb 500 kilometres up. The X-rays hurt, but electrons trapped in Earth's magnetic field do the lasting damage. NEW SCIENTIST ARTICLE

Drone Dominance Phase III The Pentagon wants about 60,000 attack drones for $450 million, and only 14 companies make the production stage. Each must deliver 400 drones in about two months at $5,000 or $3,000 apiece. Even split, those prices total $240 million, and nobody explains the other $210 million. THE INNOVATION ATTORNEY POST

Drones attack three commercial vessels off the coast of Romania, all bound for Moldova Three ships got hit near Romania's Sulina channel, all headed to Moldova's Giurgiulesti port. The Insider counts at least 13 ships struck near NATO coasts since October. THE INSIDER ARTICLE

RIP to the $1.45 Trillion FY2027 Defense Budget The midterms killed the $350 billion reconciliation piece of the request, so only the $1.1 trillion base budget has a shot. Programs like the $53.6 billion Defense Autonomous Warfare Group lived almost entirely in that reconciliation money. Expect cuts, and probably another continuing resolution. MATT VALLONE ARTICLE

German spies granted sweeping new powers Germany's parliament voted 308 to 190 to let its spy agencies hack back, not just collect intelligence. The BND can now take down servers used by state-backed hackers abroad. AL JAZEERA ARTICLE

ICE Is Paying AT&T for a Phone-Record Analytics Service Linked to Hemisphere ICE paid AT&T over $2.1 million in fiscal 2025 for analyst access to a call database with a trillion records going back to 1987. It's basically Hemisphere under a new name, a program whose federal funding reportedly ended in 2024. Experts worry ICE can search it without a judge's warrant. GADGET REVIEW ARTICLE

Palmer Luckey says selling Nvidia GPUs to China works against the CCP Luckey says Beijing doesn't want Nvidia selling to Chinese companies, because it wants its own chip industry. He claims 600,000 H200s are stuck at a port with no import permit. @OVERTON_NEWS POST

AI

How the Claude Code team thinks about verification About 90% of the team's code is written and reviewed by Claude, so Addy Osmani says checks decide what a human sees. If a check still fails after three rounds, Claude stops and calls a person. CLAUDE.DEV ARTICLE

Paul Christiano on the intelligence explosion, stealing weights, and alignment Paul Christiano puts a 1 in 3 chance on superintelligence strong enough to overthrow a government within a year. He wants labs to pause algorithmic progress until they can lock down their weights. @DWARKESH_SP POST

My voice harness is my new operating system Paul Kinlan built a voice-controlled harness that runs in the browser and lets him talk to his computer. The model writes small apps on the fly, and it can hand bigger jobs to Claude Code or Codex. He admits it has no protection against prompt injection yet. PAUL KINLAN POST

Anthropic says the key to agent memory is knowing what to forget, not how much to store Agents that save everything drown in stale notes. Claude does better when it keeps only strategies and verified rules. THE DECODER ARTICLE

Which business moats will still matter after AI Plenty of companies get protected by customers being too busy to switch, compare, or cancel. Personal AI agents don't get tired, so that protection could fade. When Meta's Muse agent launched, Planet Fitness fell as much as 11% and Schwab dropped over 6%. MICHAEL JOSEPH ARTICLE

Claude Meetings I see this as an extension of Claude Tag. Anthropic's new Claude Meetings lets Claude join your Google Meet calls as a visible participant and do work during the call. It can pull numbers from connected tools and build docs live. SÉBASTIEN DUBOIS POST

TECHNOLOGY

One router, one bug. How routine upgrades caused worldwide chaos Microsoft's email and office software went dark for millions of people after an upgrade meant to add capacity. A day later, US authorities grounded every departing flight in the country for five hours after a contractor installed a router at a remote site in Arkansas. David Swan covers both. SMH ARTICLE

Apple adds new disk controls to prevent AI agents from running amok Apple's latest macOS beta now throws up a scary warning before you give an AI app Full Disk Access. The pop-up says your Mac "will not be able to protect your data" from that app. It's in macOS 27.3 beta 2, with no release date yet. ENGADGET ARTICLE

Migrating from GitHub to Codeberg Yorick Peterse moved 70+ repos off GitHub because of AI slop. It took self-hosted CI runners to make it work. YORICK PETERSE POST

Cloudflare Gives AI Agents a Browser of Their Own Kitesurf is a cloud browser that uses 3.1 to 3.8 times less processing power than Chromium. It starts a fresh session per request, but it can't stay logged in or pass some bot checks. PYMNTS ARTICLE

Chemistry is Actually Geometry You can predict how a messy reaction network behaves just by counting its parts, with no rate constants needed. Feinberg's deficiency zero theorem says a network that scores zero and is weakly reversible has exactly one stable equilibrium in each compatibility class. The author walks through it with worked examples. CASUALPHYSICSENJOYER ESSAY

Rebuilding my voice assistant pipeline Fabio Manganiello swapped openwakeword for microWakeWord and reworked speech detection. Conversations now end 1.0-1.6 seconds after the last word instead of 6-10. FABIO MANGANIELLO POST

HUMANS

AI and complements Charles Jones and Christopher Tonetti model an economy where tasks depend on each other, so the weakest one sets the pace. Automating 1% of GDP with 10x productivity adds under 1% to output. An attack that cuts that productivity tenfold drops output 68%. MARGINAL REVOLUTION POST

Generative AI and the temporary upskilling of knowledge workers BCG gave 487 consultants ChatGPT to do data science tasks outside their skills, and they beat the control group. But without ChatGPT afterward, they showed no sign of having learned anything. NATURE HUMAN BEHAVIOUR PAPER

'It's like the wild west': how AI has rocked the literary world A $2m book deal collapsed and a prize longlist lost a novel after readers ran the text through AI detectors. Publishers still won't say what's allowed, and Coral Hart used AI to write 370 novels. SOPHIE MCBAIN GUARDIAN ARTICLE

The most important sleep metric isn't hours slept Going to bed and waking up at the same times predicts death risk better than sleep duration does. People in the most regular quintile had 30% lower mortality. BIG THINK ARTICLE

Redefining the Product Manager Role Product managers using Claude Code can now build prototypes and ship small fixes themselves, so the handoff to engineering moved instead of vanishing. Agent-heavy teams are reportedly going from one PM per eight engineers to one per twenty. Engineers still sign off on anything touching auth or core data. THE INNOVATION ATTORNEY PODCAST

US has lost measles elimination status, global health authorities say The US just lost the measles elimination status it won in 2000, because outbreaks no longer die out within a year. Over 8,000 people have been sickened in two years, and 13 have died. NPR ARTICLE

Global economy to grow 2.8% in 2027 Goldman Sachs says that's barely above this year's 2.7%. Cheaper oil and AI spending offset tighter money and fading government support. The US, Japan and the UK should beat consensus, and the AI spending surge lands mostly in the first half. BUSINESS TODAY ARTICLE

Why Nobody Can Quit: The Private Welfare State as Labor Discipline Quitting is the last bargaining chip most workers have, and health insurance is taking it away. Richard Greenwald says 24% of insured workers now stay in jobs they want to leave, up from 16% in 2021. For workers with chronic conditions it's 41%. AMERICAN AFFAIRS ARTICLE

IDEAS

The Profligacy Principle Venkatesh Rao flips Occam's razor into "waste entities." When AI keeps breaking our old categories, you need lots of ways of seeing it before you know which ones matter. VENKATESH RAO ESSAY

Bad ideas that aren't bad anymore Devon Estes says his engineering gut keeps flagging ideas that now work fine with AI. His team stopped peer code review, so the human who owns the ticket reviews Claude's PR. His fix is to ask what cost made an idea bad, and whether that cost is still there. DEVON ESTES POST

Most of AI moats are just fuel Most of what people call moats in the AI age is really fuel, because it stops protecting you the moment you stop spending. adlrocha says the real moat is a loop, where each customer makes the next one easier to win. Stripe's fraud detection is his example. ADLROCHA POST

The No AI Slop Manifesto Slop is when you spend less effort making something than the reader spends understanding it. Speakers at AIE Code Summit wrote this Agile-style manifesto. DEX HORTHY GIST

J.R.R. Tolkien on Fairy Tales, Language, the Psychology of Fantasy, and Why There's No Such Thing as Writing "For Children" Tolkien thought fairy stories ended up in the nursery by accident, because adults didn't want the furniture. In his 1939 lecture he says fantasy is harder to write than realism and that it sharpens reason. Maria Popova pulls the best passages together on The Marginalian. THE MARGINALIAN ARTICLE

Flashy courtship traits tend to evolve faster than combat weapons, study finds Ornaments like feathers and mating calls changed about twice as fast as weapons like tusks and spurs across 531 species. The model says female tastes drift randomly, so the target keeps moving. PSYPOST ARTICLE

How vibes ate the world Robin James says "vibes" have replaced norms as the way power keeps itself running. She talks through it on the CUJO podcast. CULTURE JOURNALIST PODCAST

A lot of AI progress is just persistence Shmuel Berman thinks AI looks superhuman mostly because it doesn't get lazy. It keeps grinding on tedious work humans skip. SHMUEL BERMAN ESSAY

Senior engineers still decide what ships when AI writes the code Lauren (@poteto) says AI may write most of the code, but senior engineers still own quality. They should build guardrails and teach the next generation. Her analogy: everyone can cook at home, yet restaurants still exist. @POTETO POST

A hospice doctor says humanity is in its own hospice moment A former ICU and hospice doctor says our polarized, victim-versus-villain era is a species-level deathbed. He thinks we'll look back and see it was all perfect. CINDY HUGHES X POST

AI Will Not Make You Rich Jerry Neumann says AI looks like shipping containers, not microprocessors. Most of the money goes to customers, not builders. COLOSSUS ARTICLE

My Chrome extensions are just prompts Paul Kinlan built a Chrome extension studio where the whole program is a prompt.md file. A generic runtime handles the rest. AIFOC.US ESSAY

The Tocqueville Nobody Knows Michael Lind says Tocqueville got 1830s America badly wrong. The Tocqueville worth reading is the one who warned about industrial aristocracy. AMERICAN AFFAIRS ESSAY

DISCOVERY

Absolutely love this guy Anton’s videos, and I think it’s super inspiring as a creator that you can do this well by just being curious and excited about things. This is what should all be doing, in our own ways.

A prompt for coding agents: add robust, near zero-lag online multiplayer to any real-time game Someone wrote a prompt you paste into a coding agent to add multiplayer to a real-time game. Each client simulates its own player, so controls never lag, and everyone else is drawn where they are now, not in the past. A simulated-network bench checks every change. GITHUB GIST PROMPT

Write the eval before the fix: building a WhatsApp transcriber with LLM-TDD Waldo Vanderhaeghen built WhatYouSay, a WhatsApp bot that transcribes voice notes and keeps no transcripts. Scoring against 20 reference notes took cleanup from 69.9% to 86.4% word match. WALDO VANDERHAEGHEN POST

Harness & Hunt: A Hybrid, Role-Driven Agent Architecture for Autonomous Vulnerability Research A small team built a cheap bug-hunting pipeline with separate LLM roles and hard checks like ASan crashes. It found zero-days in V8 and FFmpeg using 18.1 million tokens. GREPTALKS TALK

Why I'm building my hacking agent out of tiny, boring tools Luke Stephens says don't make the model do work a plain CLI tool can do. Wrap small deterministic tools in MCP, keep their output lean, and let the model just decide what to try. DETECTIFY BLOG POST

I let Claude Code clean out my Downloads folder every Friday Mahnoor Faisal set up a weekly Claude Code task that sorts her Downloads folder at 8 PM. A test run trashed 53 items, about 766 MB, and flagged 34 exact duplicates. MAHNOOR FAISAL ARTICLE

Browser-Use 1.0.5 Every run now gets graded by your own model after it finishes. It writes up what the site blocked or rejected. BROWSER-USE RELEASE NOTES

How to Run Vision Models on OBS Live Broadcasts You can run computer vision on an OBS stream with no plugin and no GPU. Just point OBS at Roboflow's RTMP URL. ROBOFLOW BLOG

Phone Harness Phone Harness lets Claude Code, Codex, or any agent use your real phone. It taps, scrolls, and types on an iPhone or Android with no jailbreak and nothing installed on the phone. PHONE HARNESS REPO

I stopped using AI browsers and replaced them with this Claude setup Claude's desktop app has its own built-in Chromium browser pane, and it's walled off from your real browser and logins. Nolen Jonker uses it to open ten comparison tabs at once and get back a table with the fine print. It doesn't cost extra, but it counts toward your normal Claude usage. XDA ARTICLE

Claude Code gets an experimental Code Mode for working with tools Instead of calling tools one at a time, Claude writes a script that chains them together. It can also reply to you while long commands are still running. You turn it on with CLAUDE_CODE_EXPERIMENTAL_CODE_MODE=1. CLAUDEDEVS X POST

Robots.txt Guide – Best Practices For Search & AI Crawlers You can block AI training bots like GPTBot and still let search bots in. But ChatGPT-User and Perplexity-User can ignore robots.txt entirely. SEARCH ENGINE JOURNAL ARTICLE

RECOMMENDATION OF THE WEEK

I’m updating my expectation of chaos in 2027, and unfortunately it’s not the type of thing where I think I know where it’s going and so I think you should do x, y, and z to ride the wave.

I don’t know what the wave will be. Because it’ll be 237 waves, and they’ll interact with each other.

So here are some more tangible (and perhaps slightly alarming) recommendations:

  1. Be prepared to live in your house for a long period of time, kind of like Covid type situation. Like lots of water, food, basic defense, etc. Not like crazy prepper, just enough to not feel like an idiot if you don’t have it. Ideally a solar capture / storage solution too.

  2. See if you can switch / add your income to remote options. Remote work, remote contract jobs, etc. Stuff where you can do anywhere, as diversified as possible.

  3. Keep getting better at doing your work and building with AI. Full stack, from idea to MVP, to a functional product, and then getting it out into the world. The lines between these roles are going away.

APHORISM OF THE WEEK

❝

Attacks always get better; they never get worse.

Bruce Schneier

GET THE MEMBER EDITION

You’re currently receiving the STANDARD edition.

Members help this work continue. If you enjoy the newsletter, the podcast, what I put on YouTube, or any of my open-source projects on Github, I ask you to please become a member. It allows me to stay focused on learning and building and sharing. It’s like a cup of coffee or two per month.

Plus, members get numerous benefits, including:

  • 25-50% off all UL Paid Content, including the upcoming Human 3.0 / AUGMENTED ONLINE portal!

  • Access to the extraordinary UL Member Community that includes vibrant conversations with ~1,500 of the smartest and kindest people you’ll find on the internet

  • Member-only Content, such as EDC guides on tech stacks, personal productivity routines, my recommendations on Critical skills to Build Going Forward, Trend Identification and Analysis, and more…

  • Access to the Member Archive of previous Member-only content, the Book Club archive, etc.

  • Access to The UL Book Club that’s been going monthly since 2017! One of the highlights of my and many attendees’ month!

  • Access to the Monthly Member Meet-up where we talk about our routines, productivity workflows, what’s on our minds, etc.

  • Access to In-Person Events like our dinners in Vegas, San Francisco, etc.

  • And much more coming…

This is the moment to connect with others who are smart, kind, and asking the same questions we are. Where is this all going? And how do to prepare?

Join the conversation.


This newsletter is at AI Influence Level 3, from a mixture of my own manual writing, my manual story selection throughout the week, the manual curation of the stories that go into each episode, and AI-written story snippets (which often includes my own manual writing as well). I created the AIL project in 2023 because I think it’s important that creators disclose how much of their output is AI vs. manually generated. There may come a time when I think this type of labeling is no longer necessary, but personally, for myself, that time has not yet come.