UPDATES
Hey, hope you’re starting off well this week!
New blog and video on why AI isn’t a bubble. It's been nagging me that I couldn't perfectly articulate this before now. And now I think I have the cleanest way to explain the difference between AI and the .com thing.
—
TMI Radical Sharing Lab Results (which I intend to keep sharing for kinship):
Metabolic is pristine. Sugar-wise, I mean. Great numbers
Liver great
Kidneys not so good. Too much Creatine and Celsius probably
HDL a tad too low, LDL a tad too high
VO2 is lower as well since I tripled my effort on building Kai last several months
Too many 14-hour days in the chair, less weights, less table tennis, less rucking and walking
Need to get back on it!
If you’re in the same sich, you do the same!
What I’m doing to adjust:
Stand-up alarms for standing desk so I’m not sitting all day
Foam seat pad with cutout so I’m not sitting on my junk all day
Mandatory cardio twice a week (table tennis or rucking)
Weights at least twice a week
More coffee to balance the Celsius
Exercise to help HDL (above)
LMK if you’re on a similar journey.
—
🤖 Kai Update! (he hates that emoji)
MASSIVE CONTEXT LOADING BREAKTHROUGH!
Just had the sickest idea about improving my dynamic context loading system for Kai (Claude Code). And it works!!!!
I talked before about my dynamic context loading system, which was a TypeScript file launched from the user prompt hook. But it was still a bit kludgy because I was editing actual code, and it just wasn't as easy to work with.
I decided to try loading a Claude Code command in the hook so that I could do things dynamically with context and using AI for the routing as opposed to keyword matching. But hooks can only load executable files. So I now have a
load-dynamic-context.tshook that calls aload-dynamic-context.mdfile in my commands directory. lol
It works absolutely brilliantly. I just love Markdown and how extensible this whole system is because of how commands combine prompts with executable code.
So the result of this is now my AGENTS.md files in my various projects are mostly empty! The user prompt dynamically loads the context that I need at that moment, and not a moment before!
—
Speaking of Kai PAI, I’m open-sourcing the whole scaffolding system I’m using.
So not only will I be putting the overall structure I’m using in this repo, but as I release videos like my recent video on PAI, I’ll be including code and files here in this repo.
You can ⭐ it here to follow updates. THE PAI REPO
Designed the logo myself (do you see the PAI?)
—
Just thought of a cool idea and implemented it as a command, and now it's my first commit to the repo.
UPDATE: I’ve added tons of stuff in there now! THE PAI REPO
capture-learning.mdcapture-learning.ts
Any time we learn something cool or fix something, I have Kai invoke this command and he will log to the context directory so we have a running list of lessons learned. We could then go through them and update our documentation or code based on that!
—
Switched back to Chrome since Arc/Dia got bought by Jira basically. But I will be watching their progress very closely…
Sponsor
Fortify Your Google Workspace, from Gmail to Drive.
Protect the email, files, and accounts within Google Workspace from every angle. Material Security unifies advanced threat detection, data loss prevention, and rapid response within a single automated platform so your lean team can do more with less.
Deploy in minutes, integrate with your SIEM, and let “set-it-and-forget-it” automation run 24/7. Gain multi-tenant visibility with enterprise-grade security without enterprise overhead.
CYBERSECURITY
npm’s most-used UI packages got backdoored to hijack crypto
Aikido caught malicious updates to debug, chalk, and 16 other major npm packages that inject a browser interceptor to rewrite crypto addresses and swap approvals. Supply chain keeps hitting hard. AIKIDO INCIDENT POST | MAINTAINER BLUESKY CONFIRMATION | MAINTAINER BLUESKY PHISH SHARE | HACKER NEWS COMMENT THREAD
My buddy Clint Gibler and team found tons of vulns with a 10-line Claude Code Prompt
Semgrep found legit bugs in real web apps, but with a lot of false positives. I think the project is a total success because it's probably fairly easy to clean up those false positives. The fact that it was done with nothing but a 10-line prompt is crazy. Keep in mind: Actual confirmed vulnerabilities in real code on the internet. SEMGREP BLOG POST | TECH BEHIND ASSISTANT | PROMPTFOO EVALUATION POST | 96% TRIAGE AGREEMENT POST | CLAUDE SECURITY-REVIEW ANNOUNCEMENT | CLAUDE SECURITY-REVIEW PROMPT
Hiring fraud flips zero trust from network to identity
David van Heerden lays out how fake remote hires bypass email defenses entirely, pushing teams toward zero standing privileges over castle-and-moat lockdowns. They don't need to phish you if you just hire them and they go through onboarding. THE HACKER NEWS ARTICLE | CROWDSTRIKE 2025 THREAT HUNTING REPORT | ZERO STANDING PRIVILEGES EXPLAINER – BEYONDTRUST | DEEPFAKE THREATS OVERVIEW – BEYONDTRUST | BEYONDTRUST ENTITLE PRODUCT PAGE | FREE IDENTITY RED-TEAM ASSESSMENT
Chrome gets critical fixes for code execution, patch now
Center for Internet Security says multiple Chrome bugs can lead to code execution; update to 140.0.7339.80+ now—no active exploits reported yet. CIS ADVISORY | GOOGLE STABLE CHANNEL UPDATE | CVE-2025-9864 | CVE-2025-9865 | CVE-2025-9866 | CVE-2025-9867
Scattered Spider now targets browsers as enterprises move operations there
The Hacker News explains how Scattered Spider has evolved to focus on browser-based attacks, stealing credentials through JavaScript tricks and session tokens while bypassing MFA and traditional security tools. THE HACKER NEWS ARTICLE | SERAPHIC SECURITY DEMO
Attackers can serve poisoned websites only to AI agents, not humans
Shaked Zychlinski discovered attackers can fingerprint AI agents and serve them completely different malicious websites than what humans see. ZYCHLINSKI'S RESEARCH PAPER
US puts $10 million bounties on three Russian FSB hackers
The State Department is offering $10 million for info on three FSB officers who hacked over 500 energy companies across 135 countries. TOM'S HARDWARE ARTICLE | REWARDS FOR JUSTICE POST | DOJ INDICTMENT DETAILS | REWARDS FOR JUSTICE SITE
Sponsor
AI Agents That Investigate Like Engineers
Maze trains AI to analyze vulnerabilities the way humans do: testing exploit scenarios against your actual environment and controls. The outcome: 80–90% of false positives cleared, only real risks prioritized, and fixes routed directly to the right owners. Built by ex-Meta, Elastic, Amazon, and GitHub engineers.
CISOs face growing pressure to hide breaches CSO ONLINE ARTICLE
CISA pushes universal software ingredient lists CYBERSCOOP ARTICLE
You can find bugs by reading code ALEX'S BLOG POST
Over 1,100 exposed Ollama servers found via Shodan CISCO SECURITY ARTICLE
NATIONAL SECURITY
U.S. launches a Department of War site U.S. DEPARTMENT OF WAR SITE
America steps back from the information fight
Anne Applebaum says the Trump team is gutting US overseas media and anti-propaganda programs, basically handing China and Russia full reign to control the narrative in places we used to have influence. THE ATLANTIC ARTICLE | VOA FIRINGS POLITICO REPORT | JUDGE LAMBERTH RFE/RL ORDER NYT | WASHINGTON POST ON VOA DIRECTOR RULING | WSJ ON CHINA FILLING VOA SLOTS | WASHINGTON POST ON VOA IRAN SCRAMBLE
F-35s head to Puerto Rico as U.S.-Venezuela standoff heats up
Howard Altman says the U.S. is sending 10 F-35s to Puerto Rico after Venezuelan F-16s buzzed a Navy destroyer and Trump greenlit a lethal strike on a cartel boat. This guy has a funny way of showing how isolationist he is. THE WAR ZONE STORY | REUTERS CONFIRMATION | VENEZUELAN F-16S NEAR DESTROYER | TRUMP VIDEO OF BOAT STRIKE | NYT LEGAL QUESTIONS | F-35 ELINT PRIMER
State Department tightens where you can apply for US visas
The State Department says nonimmigrant visa applicants now need to interview in their country of nationality or residence. STATE DEPARTMENT VISA POLICY UPDATE | GLOBAL VISA WAIT TIMES | U.S. EMBASSY AND CONSULATE SITES
Sudan closes oil facilities after drone strikes OODALOOP COVERAGE
DHS wants $100 million for counter-drone systems to protect America
The Department of Homeland Security plans to spend over $100 million on anti-drone technology between 2026 and 2030 to defend against hostile drones threatening critical infrastructure and public safety. DHS ACQUISITION NOTICE | REGISTER ARTICLE | PROJECT FLYTRAP | ANTI-DRONE RIFLE | TRUCK-MOUNTED LASER UNITS
China is using its private sector to advance military AI capabilities
Researchers at Shanghai Jiao Tong University developed AI for automated naval "kill webs" that adapt in real time during combat, and six days later China's military awarded them a defense contract. But I mean, in fairness, we're doing the same thing. OODALOOP ARTICLE
AI
OpenAI launches an academy and jobs board to reskill displaced workers
The Register’s reporting that Fidji Simo is rolling out OpenAI Academy plus a jobs board—with Walmart as an anchor partner—to retrain folks AI displaces and place them with OpenAI customers. THE REGISTER ARTICLE | OPENAI BLOG—FIDJI SIMO | OPENAI ACADEMY | GEORGIA TECH COLLABORATION | WHITE HOUSE “NO WOKE AI” CONTEXT | THE HILL DINNER PIECE
Anthropic agrees to pay authors at least $1.5B in settlement
Wired’s Andy Greenberg says Anthropic will pay at least $1.5B to authors—roughly $3K per book—instead of risking a trial that some experts thought could land near a trillion. WIRED ARTICLE | EARLIER WIRED COVERAGE
Anthropic valued at $183B after $13B raise THE VERGE STORY
Anthropic blocks Chinese-owned firms from Claude TOM'S HARDWARE STORY
DeepMind's AI cuts gravitational wave detector noise 100x DEEPMIND BLOG POST
AI agents could reshape entire economies within the next decade
Economists examine how autonomous AI agents that plan and execute complex tasks without human oversight might transform markets, organizations, and institutional structures. ARXIV PAPER
AGENTS.md emerges as the universal standard for AI coding agent instructions
Developers appear to be consolidating around AGENTS.md as the single format for instructing AI coding agents, replacing the chaos of tool-specific files like CLAUDE.md and GEMINI.md. I’ve renamed all mine to AGENTS.md within Claude Code. SOCKET DEV ARTICLE | AGENTS.MD PROJECT | GITHUB COPILOT ANNOUNCEMENT | GITHUB SEARCH RESULTS
High school senior says AI is destroying real learning for his generation
A New York high school senior describes how classmates constantly use ChatGPT for everything from literature annotations to math homework, turning education into copy-paste exercises instead of actual thinking. THE ATLANTIC ARTICLE
AI is destroying education for high schoolers THE ATLANTIC ARTICLE
Evidence shows AI is already replacing entry-level workers
Derek Thompson analyzes new data showing AI is hitting young workers hardest—companies are cutting junior roles while keeping senior staff who manage the AI systems. DEREK'S AI JOBS ANALYSIS | HACKER NEWS DISCUSSION
TECHNOLOGY
Google gets to keep Chrome but must end exclusive deals and share search data CNBC BREAKING NEWS | ORIGINAL DOJ CASE | AUGUST 2024 RULING | GOOGLE APPEAL ANNOUNCEMENT
OpenAI moves to make its own AI chips with Broadcom
OODAloop says OpenAI’s building its own chips with Broadcom next year to cut Nvidia dependence and feed model hunger. OODA LOOP STORY | OODA LOOP HOMEPAGE
UC Santa Cruz engineers show WiFi signals can measure heart rate
UC Santa Cruz researchers demonstrated that household WiFi devices can accurately measure heart rate without any wearables, using $5-10 ESP32 chips and machine learning to detect tiny signal variations from heartbeats up to 10 feet away. UC SANTA CRUZ ARTICLE | IEEE STUDY
SpaceX can now launch 120 times a year from Florida TECHCRUNCH ARTICLE
HUMANS
Jobs growth stalls as unemployment hits a four-year high
CNN’s August jobs piece says we added just 22,000 jobs and unemployment’s now the highest in four years, with revisions and weak openings signaling a slowdown. CNN JOBS REPORT | JULY JOBS REPORT CONTEXT | JOLTS: FEWER OPENINGS THAN SEEKERS | BLS SUSPENSION CONTROVERSY | TRUMP FIRES BLS COMMISSIONER | ANTONI POLITICIZATION CONCERNS
Clopidogrel may outperform aspirin for heart attack prevention WIRED STORY
Private equity rentals make suburbs more diverse but apply pressure to buyers
Alex Mayyasi shows the weird trade-off: corporate landlords raise prices for would-be buyers, but their single-family rentals open up better neighborhoods to lower-income, younger, and nonwhite families. NPR PLANET MONEY STORY | CHANG PAPER ON DIVERSIFYING SUBURBS | FED PHILLY INSTITUTIONAL INVESTOR STUDY | NORTH CAROLINA SCHOOL ACCESS STUDY | REDFIN INVESTOR PURCHASE DATA | BLOOMBERG MAGNETAR TAX CASE
More people are opting out of news to protect their mood
The Guardian’s reporting shows a big jump in “news avoidance,” with the Reuters Institute finding 40% now often skip news—mostly to avoid anxiety and nonstop negativity. Indeed. GUARDIAN INTERACTIVE STORY | REUTERS INSTITUTE REPORT SUMMARY | MDPI STUDY ON NEWS EXPOSURE | NATURE PAPER ON DISTRESSING EVENTS | PNAS PAPER ON GRAPHIC IMAGERY | APA STRESS IN AMERICA
Mathematicians just broke an 87-year-old knot theory conjecture SCIENTIFIC AMERICAN ARTICLE
Attention loops quietly rewrite your model of reality
Henrik Karlsson says whatever you stare at long enough starts echoing back and reshaping you—so pick loops that compound insight, not your neuroses. HENRIK KARLSSON ARTICLE | HACKER NEWS DISCUSSION
Phones on the toilet raise hemorrhoid risk NBC NEWS COVERAGE
IDEAS
The quality of a government is determined by the ease with which its population can be emotionally manipulated. Emotion vs. Critical Thinking. Bro knowledge vs deeper education. Tribalism vs. Humanism. It's not about religion vs. reason as I used to think when I was younger. It's about whether you deeply understand how the world works and can identify and untangle attempts to manipulate you emotionally. When I think about our government for the last 5-10 years, I'm starting to think less about all the messed-up things that the government is doing, and instead about the deterioration of our education, critical thinking, and wisdom as a population that allowed it. It's like trying to fill up a bathtub when you've taken the plug out of the bottom. And that's why I think the world, and especially Europe, is going to turn away from us. The problem isn't who we've elected. It's the fact that there's no reason to think we won't do it again.
DISCOVERY
Rick Rubin’s The Way of Code
Didn’t know he was a coder. Site design is epic though. Checking it out. 👀 THE WAY OF CODE
Muscle-mem turns repeated agent tasks into deterministic replays
Erik (pig-dot-dev) built a Python “behavior cache” that records agent tool-call sequences and safely replays them, skipping the LLM for repeats and falling back when checks fail. Basically, it looks at all the different steps that your tool takes to accomplish something and remembers that. Really cool, I'm gonna mess with this. MUSCLE-MEM GITHUB | MUSCLE MEM DISCORD
Being good isn’t enough to win anymore
Josh Swords argues that craft alone doesn’t carry you—you need distribution, timing, and taste, or your “good” never gets found. This is why I'm arguing that everyone needs to become a creator. It's not that they need to become what a creator is today. Creator-ing just needs to be the new way of living. Thinking, building, sharing. I'm arguing that these are core human things, not creator things. BEARBLOG ARTICLE | HACKER NEWS DISCUSSION
Blogs used to be weird and personal JETGIRL ARTICLE
Cutting e2e test time 84% with Claude Code SDK
Jampa Uchoa shows how the Claude Code SDK cuts flaky end-to-end test time by 84% with a smart hybrid approach. SUBSTACK POST
Make your own handwriting font easily MAKING A FONT OF MY HANDWRITING ARTICLE
Claude fakes tool use unless you enforce strict toolcalling. I've seen this. That's why I'm very explicit with all my agents and context loading. REDDIT POST
AI enables live phone calls across language barriers BBC TECH LIFE EPISODE
Tech interviews reward complexity over practical engineering DIALLO'S BLOG POST
Chibi analyzes why users churn CHIBI HOMEPAGE
Random walks almost never return home in 3+ dimensions NOLTE'S ANALYSIS
RECOMMENDATION OF THE WEEK
I'm going to reuse one that I typically have to ask myself, and I think is a really good question to be asking ourselves at this moment of extraordinary change.
What would you do if you were not afraid?
Normally, I would follow that up with "You shouldn't be afraid, because everything is ok.” But everything isn't okay.
But it could be that right now is the exact best time for reinvention.
So the question still stands.
APHORISM OF THE WEEK
The privilege of a lifetime is to become who you truly are.
GET THE MEMBER EDITION
You’re currently receiving the STANDARD edition.
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.