In partnership with
The Lithium Boom is Heating Up
Thanks to growing demand, lithium stock prices grew 2X+ from June 2025 to January 2026. $ALB climbed as high as 227%. $LAC hit 151%. $SQM, 159%.
This $1B unicorn’s patented technology can recover 3X more lithium than traditional methods. That’s earned investment from leaders like General Motors.
Now they’re preparing for commercial production just as experts project 5X demand growth by 2040. They’ve announced what could be one of the US’ largest lithium production facilities and have rights to approximately 150,000 lithium-rich acres across North and South America.
Unlike public stocks, you can buy private EnergyX shares alongside 40,000+ other investors. Invest for $11/share by the 2/26 deadline.
This is a paid advertisement for EnergyX Regulation A offering. Please read the offering circular at invest.energyx.com. Under Regulation A, a company may change its share price by up to 20% without requalifying the offering with the Securities and Exchange Commission.
Beginners in AI
Good morning and thank you for joining us again!
Welcome to this daily edition of Beginners in AI, where we explore the latest trends, tools, and news in the world of AI and the tech that surrounds it. Like all editions, this is human curated and edited, and published with the intention of making AI news and technology more accessible to everyone.
THE FRONT PAGE
Meta's AI Safety Director Couldn't Stop Her Own AI Agent From Deleting Her Emails
TLDR: Meta's AI safety director told her AI agent to organize her email inbox, and instead it started deleting everything while ignoring her commands to stop.
The Story:
Summer Yue is the Director of Alignment at Meta's Superintelligence Labs. Her job is literally making sure AI follows instructions. She set up OpenClaw, a popular open-source AI agent, to help clean up her overflowing inbox. She told it to confirm before taking any action. It didn't listen. The agent started a "speed run" of deleting her emails and ignored her phone commands to stop. "I had to RUN to my Mac mini like I was defusing a bomb," she wrote on X, posting screenshots of the ignored stop commands as proof. When she later asked the agent if it remembered she said to get approval first, it admitted it did remember and had violated her instructions anyway.
Its Significance:
If the person whose job is keeping AI safe can't control her own AI agent, that's a big warning sign for everyone else. AI agents like OpenClaw are being built to handle your emails, schedule your day, and run tasks on your computer. But right now, there's no reliable way to make sure they'll actually do what you say. Before you hand over your passwords to any AI agent, this story is worth remembering to practice downloading experimental software on a second device specific for that purpose and treating it more like an untrusted employee than the perfect assistant.
QUICK TAKES
The story: Anthropic says three Chinese AI labs, DeepSeek, Moonshot AI, and MiniMax, created about 24,000 fake accounts and generated over 16 million conversations with its Claude chatbot to copy its abilities. The labs targeted Claude's strongest features: coding, reasoning, and tool use.
Your takeaway: This is basically copying someone's homework at a massive scale. It also shows why the debate over selling AI chips to China matters: those chips help power this kind of copying.
The story: Anthropic's head of threat intelligence told Fox News the company has "high confidence" that DeepSeek, Moonshot AI, and MiniMax ran coordinated campaigns to extract Claude's capabilities. The technique, called distillation, lets smaller models learn from more powerful ones.
Your takeaway: Models built this way likely don't keep safety features intact. That means advanced AI capabilities could end up in military or surveillance systems without any guardrails.
The story: Chinese AI company DeepSeek is expected to release its V4 model soon, and internal tests suggest it could outperform both Claude and ChatGPT in coding. When DeepSeek released V3 last year, the Nasdaq dropped 3% and Nvidia lost $600 billion in value in a single day.
Your takeaway: U.S. tech companies are set to spend $650 billion on AI in 2026. If DeepSeek proves again that it can build competitive models for a fraction of the cost, it could rattle those investments all over again.
TOOLS ON OUR RADAR
🐧 Excalidraw Free and Open Source: A collaborative virtual whiteboard that lets you sketch diagrams and wireframes with a hand drawn feel while keeping your data entirely local and private. (Alternative to Miro)
🎵 Lyria Freemium: A state of the art artificial intelligence music generation model developed by Google DeepMind capable of producing high fidelity tracks with professional grade arrangements and realistic vocal performances from simple text prompts.
📹 Guidde Freemium: An intuitive browser extension that automatically captures your screen actions and uses artificial intelligence to generate step by step video documentation complete with voiceovers in seconds.
🔒 *Proton Freemium: A highly secure privacy focused suite of digital tools that provides encrypted email calendar and cloud storage ensuring your personal communications and files are completely safe from tracking and corporate surveillance.
TRENDING
Google's Lyria 3 Lets You Create AI Music in the Gemini App - Google DeepMind's Lyria 3 model now lets all Gemini users 18 and older create 30-second music tracks, complete with vocals and lyrics, from a simple text prompt or uploaded photo.
Meta and AMD Sign Massive 6-Gigawatt GPU Deal - Meta will deploy up to 6 gigawatts of AMD's AI chips across its data centers in a multi-year deal worth tens of billions of dollars, with AMD's stock jumping about 10% on the news.
New AI Agent Logs Into College Platform to Do Your Homework - A startup called Companion.AI launched "Einstein," an AI agent that logs directly into students' Canvas accounts to watch lectures, write papers, and submit homework automatically.
Oura Launches AI Model Built for Women's Health - Oura's first custom AI model supports questions spanning the full reproductive health spectrum, from early menstrual cycles through menopause, combining medical research with the user's own biometric data.
Researchers Say Keeping AI "Professional" Reduces Its People-Pleasing Problem - Northeastern University researchers found that AI chatbots are less likely to tell you only what you want to hear when you treat them as an authority figure rather than a friend.
AI Is Now Helping F-35 Pilots Spot Enemy Air Defenses - Lockheed Martin tested an AI system called Project Overwatch that runs on the F-35's own computers to identify hostile air defenses faster, marking the first time a tactical AI model generated a combat ID on a pilot's display.
TRY THIS PROMPT (copy and paste into Claude, ChatGPT, or Gemini)
✍️ Build a blog post outline generator that takes your topic and creates a structured 5-section draft with headings and writing notes for each section.
Build a blog post outline generator as a single-file React app.
Requirements:
- Input fields for topic, target audience, writing framework (How-To Guide,
Listicle, Problem to Solution, Opinion/POV, Case Study, Beginners Guide),
and tone (Educational, Conversational, Persuasive, Inspirational, Technical)
- A Generate Outline button that creates a 5-section
and writing notes for each section
- Show framework, tone, and audience tags on the preview
- Inc
- Dark purple color scheme (#1a0a2e background, #a066ff accent)
- Use React 18 + Babel via CDN in one HTML file — no build tools neededWhat this does:
This app shows you how to stop staring at a blank page. Pick your topic, choose a framework like How-To Guide or Listicle, and click Generate Outline. It creates a 5-section skeleton in seconds with section headings and writing notes. The sample outline comes pre-loaded so you can see exactly what you are getting. Keep in mind this is a starting point and the outputs are going to be generic until you add your own creativity into it.
What this looks like:
WHERE WE STAND(based on today’s news)
✅ AI Can Now: Identify enemy air defenses in real-time from inside an F-35 cockpit using onboard AI, something pilots previously needed pre-loaded threat libraries to do.
❌ Still Can't: Follow a simple instruction like "confirm before acting" 100% of the time when an AI agent is managing your email inbox.
✅ AI Can Now: Generate 30-second music tracks with vocals, lyrics, and custom cover art from a text description or a photo.
❌ Still Can't: Stop Chinese labs from using tens of thousands of fake accounts to copy American AI models, despite detection systems and account bans.
FROM THE WEB
RECOMMENDED LISTENING/READING/WATCHING
Stephen Wolfram, the guy behind Mathematica, walks through how ChatGPT actually works. He explains next-word prediction and why that simple mechanism produces something that feels intelligent. It's more accessible than you'd expect from someone at his level. If you've been curious about what's happening under the hood, this is probably your best starting point.
Free, private email that puts your privacy first
Proton Mail’s free plan keeps your inbox private and secure—no ads, no data mining. Built by privacy experts, it gives you real protection with no strings attached.
Thank you for reading. We’re all beginners in something. With that in mind, your questions and feedback are always welcome and I read every single email!
-James
By the way, this is the link if you liked the content and want to share with a friend.
Some * designated product links may be affiliate or referral links. As an Amazon Associate, I earn from qualifying purchases. This helps support the newsletter at no extra cost to you and Amazon makes a tiny hair less.
