AI articles from X
Discover curated AI articles from X about LLMs, models, infrastructure, AI strategy, and the business impact of artificial intelligence.
80
Articles
Top contributors in AI
Expand to see who leads this category by article count and engagement.
Top contributors in AI
Expand to see who leads this category by article count and engagement.
Ranked from the current article set in this category using article count and engagement totals.
No Defaults
Defaults reduce friction, conserve energy, and let us move fast. But we rarely stop to ask whether they still make sense. -- Every time I land in the UK, I catch myself hesitating at the curb. Look
17 Claude Skills → $500/Day → $10K/Month. FULL COURSE
Six months ago I was figuring out how to make Claude write better emails. Today those same skills are generating $312 a day. Compounding. Every day. Not from one client. Not from one product. From 17
How to Make Knowledge Graphs Blazing Fast
So, you have built a knowledge graph. It has millions of nodes, hundreds of edge types, and a pile of triples that would make any data engineer proud. Then someone asks a perfectly reasonable
AI Knowledge Layer (and why your agents are useless without it)
this article gives you a two-layer system that makes every AI agent you run smarter. 20 minute setup, gets better every day, fully open source. karpathy recently talked about shifting most of his
How to get Rich and Healthy, without getting lucky
I defeated Lyme and Long Covid with peptides. I went long AMD at $4.2, Tesla at $14 and Palantir at $7. I didn't get lucky. I had a framework. Most people think wealth and health are separate games
Proof of Abundance… And How to Survive It
We forget the world is getting better on so many levels… Back before “Abundance” was a buzzword, in 2012, I co-authored the NY Times bestselling book Abundance: The Future Is Better Than You Think.
7 Claude Skills That Sell for $2,000-$5,000/Month. I Built All 7. Here's the Exact Blueprint
Three weeks ago, I built a Claude Skill in 47 minutes. It takes a 45-minute podcast episode and turns it into 15 pieces of platform-specific content — tweets, LinkedIn posts, email newsletters,
Is Apple pretending to lose the AI race?
If you think Apple is falling behind in AI, you need to read this. Because the numbers say otherwise... Four companies are expected to spend $635 billion on AI infrastructure in 2026. Apple is
What if personal AI was less iMessage and more Apple Notes?
Last week I gave a talk about some of the design ideas in Lightpage at a @resonantcompute event. Lightpage started with a simple question: what if our primary interface for personal AI was more like
LLMs Are Not Brains, and Here Is What Is Actually Missing for AGI
Today I am writing something that has been bothering me for months. This is a long format essay with my true feelings and analysis. Recently I have AI psychosis and extreme anxiety, as I use a lot of
Are Developers Really Going Away by 2030?
For years, the story sounded simple: AI would become one giant assistant, and everyone would just talk to it. But inside real companies, the picture is different. Microsoft’s 2024 Work Trend Index
200 Powerful Claude Prompts
That Replace Hours of Manual Work Category 1: Coding & Debugging Prompts 1–50 1. Code Review Review this code as a senior engineer. Identify bugs, security vulnerabilities, performance issues, and
120B Parameters on $70K Hardware Lost to 27B on a $900 GPU. Here's the Data.
i tested two $70K datacenter flagships head to head then compared the same test on a $900 consumer GPU. the results broke how i pick models for autonomous coding. everything is reproducible. the
Turn Claude into your personal financial coach (full guide)
Turn Claude into your personal financial coach (full guide) Most people pay $150 an hour for a financial advisor who meets them twice a year and tells them to diversify. You're making the same
What AI Can’t Kill
The four layers that survive every model update. The day I launched Hyperfocal, Adobe announced Firefly inside Lightroom. Adobe just showed the world they’re building prompt-based photo editing
The Claude Code Engine, Nobody Understands
Most developers treat it like autocomplete. Underneath, it's an autonomous agent runtime with features 90% of users have never touched. I went through the full source-level leaked architecture and
How to finally trust Claude's advice (using Karpathy's LLM Council method)
Claude just tells you what you want to hear. Every time. You can’t trust it. So I built a skill that forces 5 AI advisors to argue about your question, anonymously review each other's work, and hand
The Only Moats That Matter
In a world where AI can build any software, replicate any product, and automate any process, what actually makes a business defensible? Most of the answers I hear are wrong. They're all bets on the
Claude Code vs Codex: One Clear Winner
If you're building software with AI in 2026, you're choosing between two tools. Claude Code or Codex. I chose wrong at first. Then I switched. This is what I learned. I'm 17. My brother Logan and I
How I turned OpenClaw into a wikilink skill-graph (full walkthrough)
One of the things I've hated the most about OpenClaw is the inefficiency to scale it past a couple of interesting skills. Yes you can setup cron jobs. Yes you can setup multiple agents. But the
The Gravity of Winners
A Brief History of How Value Learns to Concentrate, and What Comes Next. There is a pattern running through the entire history of commerce that is obvious in retrospect and nearly invisible in the
How to Use Claude Dispatch to Run Your AI While You Sleep
The biggest complaint about Claude Cowork since it launched in January has been five words: I have to be there. You had to sit at your desk. You had to keep the app open. You had to watch Claude work.
I don't understand AI, yet I built a million dollar business with it
Three years ago I was in Morocco with my 16-year-old brother and $500 a month... and today I run an AI film studio that works with billion-dollar companies, and governments are putting our films into
How To Reason About A Messy Future
Introduction The first time I realized we were heading towards an inflection point was when I heard the music slowing down at my previous role, even as everyone around me pretended nothing would
Everyone using AI has about 12 months to develop these 3 moats
Right now 90% of people using AI are doing the same thing: prompting, accepting, shipping. no judgment, no taste. that means if you develop even one of these three moats, you stand out immediately.
Getting started with Codex: Best practices for better results
If you’re new to @OpenAI Codex or coding agents in general, this guide will help you get better results faster. It covers the core habits that make Codex more effective across the CLI, IDE extensions,
When code is free, research is all that matters
The most important people of this new era won't be engineers; they'll be researchers. When anyone can build for free, the differentiator is knowing what's worth building and whether it’s buildable at
How to train the best embedding model in the world
An unexpected side effect of graduate school was that I became the best person in the world at training text embedding models. Midway through my PhD, I trained a state-of-the-art model (using a
The Retail Barbell Effect: Why the Middle is Dying
Why brands must win at AI-driven discovery and real-world experience or get crushed in the middle I believe we’re moving toward a barbell strategy in marketing: extremely digital on one side and
Elon and the Seven Doers
Once upon a time - not in a forest long ago, but in a restless age of circuits and steel - there lived a hero named Elon. He was not born in a castle, nor raised among princes. He came instead from a
Costless Sacrifice
Originally on Not Boring The Old Testament’s 2 Samuel tells the tale of King David as he unifies the tribes of Israel and establishes Jerusalem as the nation’s capital. The world was slower then,
Agency is the Last Moat
Four days ago Jack Dorsey fired 4,000 people. Nearly half of Block's entire workforce. Gone. His stock jumped 18%. His explanation was five words: "Intelligence tools have changed everything." Last
![how i dealt with AI anxiety as a non-AI founder [self-help edition]](https://pbs.twimg.com/media/HCKMenRaoAA0nVj.jpg)
how i dealt with AI anxiety as a non-AI founder [self-help edition]
If you are still in AI denial, close this article. You are not my audience. Step One: stop binge reading I’ve worked in technology for fifteen years, and I’ve learned that leaning in early, instead of
How to set up OpenClaw Agents that actually get better Over Time (My exact stack after 40 Days)
My agents get smarter every day. All I do is talk to them. Not tweak prompts. Not swap models. Not rebuild the architecture. Just talk. Give feedback. Watch them write it down. 40 days ago, my content
How AI Is Changing the Workforce
I’m incredibly bullish on AI. In fact, I truly believe that if you’re not setting aside at least one hour every day to explore, learn, ask questions and utilize this rapidly developing technology,
Lessons from Building Claude Code: Seeing like an Agent
One of the hardest parts of building an agent harness is constructing its action space. Claude acts through Tool Calling, but there are a number of ways tools can be constructed in the Claude API with
Why New Developers Should Avoid Pure Vibe-Coding for App Development
AI tools like Rork, Replit, and other vibe-coding platforms make app development feel almost magical. You describe what you want, the tool generates code, and suddenly you have a working app. For
how to turn nano banana pro + AI UGC + viral IG reels into a $100M cash machine in 2026
there's a specific production method allowing brands to generate 100+ pieces of user-generated content weekly at costs approaching zero while maintaining visual quality indistinguishable from real
Extended Overthinking
I’m tired of AI lately. Not of what it can do. Of what it does to me. Every new model, every new tool, every new workflow. Yes, they’re powerful. Yes, they let us do more. But there’s this weird
The Trust Layer
The piping was never the hard part. We keep handing AI more access and more autonomy without asking how much of that trust is actually earned. The tools are getting incredibly good, but "incredibly
Your Org Structure Is My Opportunity
A founder told me that his hardest problem isn't the product or the customers. It's hiring. He can't find people who can do more than one thing well. And I think I know why. A generation of talent
Elon Musk conversation with Grok about the meaning of life
Elon Musk shared a rare conversation with the public that he had with Grok about the meaning of life. I'm sharing it here because his questions and probing are important and may also help you in your
The Future Cannot Be Paused: Ben Horowitz on AI, Capital & the Transformation Ahead
If you’re building a company, investing capital, or trying to understand where the world is headed in the next 3-5 years, this Newsletter is FOR YOU! So much going on… This week, on Moonshots we sat
Addition Under Pressure
I asked Claude Code and Codex to each train the smallest possible transformer that can do 10-digit addition. Claude Code came back with a 6,080-parameter model and Codex came back with 1,644
Against “Against Taste”
Taste did not suddenly arrive. Every few years the word comes back into fashion and starts carrying more weight than it should. But taste itself is simple. It is knowing what is good. That sounds
i benchmarked K2.5 (1T params) on 1x to 8x 3090s. every result, every crash, every dollar.
a single RTX 3090 can run a 1 trillion parameter model. a GPU you can find used for $800. the same card people game on just ran the largest open-weight model ever released. i wanted to know how far
I'm Scared.
This is a feeling that is very hard to put into words. Even sitting here writing this now, I'm still unsure of how to articulate this because it's hard for me to wrap my head around. But I'm going to
How To Generate AI Influencers That Actually Look Real
How To Generate AI Influencers That Actually Look Real If your AI Influencers still look fake or “too perfect”, it’s not the model’s fault. It’s your prompt. AI models like Grok Imagine, Kling 3.0,