Sponsored by
Do your searches always hit dead ends?
Nearly half of users abandon a search without getting the result they wanted. Instead, they’re stuck in a loop of irrelevant results, slow-to-load articles and contradicting advice.
heywa is a whole new way of searching. It gives your result as visual & concise stories, meaning you get answers at a glance.
And if you want to explore your topic further, you can tap through your search journey without having to re-prompt and start again.
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
China Kills Meta AI Deal Worth 2 Billion For Manus Software
TLDR: China just ordered Meta to undo its $2 billion purchase of AI startup Manus, even though the deal was already done and 100 workers had moved into Meta's Singapore office.
The Story: On Monday, China's top economic planner told Meta and Manus to walk back the acquisition. Manus makes AI agents, software that can do tasks on its own like research, coding, and data analysis. Meta announced the $2 billion buy back in December 2025 and folded the team in fast. Around 100 Manus workers moved into Meta's Singapore office, and the CEO Xiao Hong now reports to Meta's COO. There's just one problem: Hong and the chief scientist are stuck in mainland China under exit bans and can't leave. Manus was started in China in 2022, then moved its headquarters to Singapore in 2025 before Meta scooped it up. Beijing says the deal violates its rules on foreign investment in sensitive tech. Meta said the purchase "complied fully with applicable law" and expects "an appropriate resolution."
Its Significance: This is one of the boldest moves China has made to keep AI talent and tech inside its borders, and it lands weeks before a planned summit between President Trump and Xi Jinping. For Meta, it's a blow to the plan to catch up to OpenAI and Google in AI agents. For everyone else, it's a warning shot: any startup with Chinese roots, even one that moved overseas, may not be safe to buy. Expect more deals to fall apart, more US scrutiny of Chinese-linked AI firms, and a faster split between American and Chinese AI as both sides try to lock in their tech.
QUICK TAKES
The story: Analyst Ming-Chi Kuo says OpenAI is working on its own smartphone with chip partners MediaTek and Qualcomm, plus Luxshare for assembly. The twist: instead of opening apps, you'd ask one AI agent to handle tasks, like booking a flight or paying a bill. Mass production is targeted for 2028.
Your takeaway: This is OpenAI trying to escape Apple and Google's app store rules so it can put AI everywhere on the phone. If it ships, it'd be the first real test of "no apps, just agents," and your phone might stop being a grid of icons and start feeling more like a personal assistant you talk to.
The story: Nature Medicine, one of the most respected medical journals on the planet, published an editorial saying there's barely any proof that AI tools actually help patients, doctors, or hospitals. A separate JAMA Medicine study found that when symptoms were unclear, top AI models got the diagnosis wrong more than 80% of the time.
Your takeaway: AI looks great in clean lab tests and falls apart in real exam rooms. If your doctor mentions using AI to help with your case, it's fair to ask which tool, what it's been tested on, and whether a human reviewed the result. The hype is way ahead of the proof.
The story: Coachella teamed up with Google DeepMind at the 2026 festival to build three AI tools, including one that records a live show and rebuilds it as a 3D space fans can walk around in. The tools use what researchers call "world models." A world model is an AI that learns how a place or environment works, then builds an interactive version you can move through, kind of like a video game level the AI made from watching reality.
Your takeaway: World models are a big deal because they let AI plan, predict, and create whole environments instead of just answering questions. If this works, smaller artists could get the same fancy stage tools as Beyoncé, and you might explore a concert from your couch the day after it happens.
TOOLS ON OUR RADAR
📻 AzuraCast Free and Open Source: A complete web radio management suite that allows you to easily start your own internet broadcasting station and manage your digital music playlists online.
✍️ Etherpad Free and Open Source: A highly customizable document editor that lets large groups of people type on the same page at the exact same time making it incredible for collaborative drafting.
🏫 BigBlueButton Free and Open Source: A specialized web conferencing system built primarily for online education that includes digital whiteboards breakout rooms and tools designed specifically for teaching.
🎞️ HandBrake Free and Open Source: An essential desktop utility that easily converts video files from almost any format into modern and widely supported file types while significantly reducing their overall size.
TRENDING
New Browser Plugin Adds Typos to Your AI-Generated Emails to Make Them Look Real - A Harvard student vibe-coded a Chrome plugin called "Sinceerly" using Claude that messes up your AI-written emails on purpose. Three modes: Subtle, Human, and CEO. The CEO mode goes all lowercase and sometimes adds "sent from my iPhone."
Tesla Quietly Buys Mysterious $2 Billion Entity - A short note in Tesla's tax filing says it agreed in April 2026 to spend up to $2 billion in stock on an unnamed AI hardware company. About $1.8 billion of that depends on hitting performance milestones. Nobody outside Tesla seems to know what the company is or what it does.
Your Former Employer Is Selling Your Slacks and Emails to Train AI - SimpleClosure, a startup that helps companies shut down, just launched Asset Hub. It lets dead companies sell off Slack messages, emails, and code to AI labs that need fresh training data. SimpleClosure has done close to 100 of these deals in the past year, with payouts from $10,000 to $100,000 per company.
Anthropic Created a Test Marketplace for Agent-on-Agent Commerce - In a test called Project Deal, 69 Anthropic employees got $100 each and let AI agents buy and sell stuff for them. The agents made 186 deals worth over $4,000. The catch: people using better AI models got better prices, and the people on the losing end didn't notice.
Apple Under Ternus: What Comes Next for the Tech Giant's Hardware Strategy - Apple named hardware chief John Ternus as Tim Cook's successor. He'll take over later this year. Expect more focus on AI-powered devices: smart glasses, AI AirPods, a foldable iPhone in September, and home robots, including a tabletop assistant with a robotic arm.
OpenAI Just Updated Its Founding Principles - Sam Altman published a new set of guiding principles. The 2018 version said AGI 12 times. The new version says it twice. The promise to stop competing if a safer rival gets close to AGI? Gone. The new doc says OpenAI may need to "trade off some empowerment for more resilience."
The Researcher Trying to Build Superintelligence Without Language Models - David Silver, the DeepMind researcher who built AlphaGo and AlphaZero, left to start Ineffable Intelligence. He thinks ChatGPT-style models are stuck because they only know what humans already know. His startup is using reinforcement learning, where AI learns by trial and error in simulations, and reportedly raised $1 billion at a $4 billion valuation.
How Jennifer Aniston’s LolaVie brand grew sales 40% with CTV ads
The DTC beauty category is crowded. To break through, Jennifer Aniston’s brand LolaVie, worked with Roku Ads Manager to easily set up, test, and optimize CTV ad creatives. The campaign helped drive a big lift in sales and customer growth, helping LolaVie break through in the crowded beauty category.
TRY THIS PROMPT (copy and paste into Claude, ChatGPT, or Gemini)
🌀Enter your age and one choice you didn't make. Read the life you could have lived.
Build me a single-file HTML app I can open in my browser without any setup. Pure vanilla HTML, CSS, and JavaScript with one fetch call to the Anthropic Claude API. Create the Parallel Life Simulator — a tool where you enter your age and one choice you didn't make, and Claude narrates the alternate timeline in literary detail.
Aesthetic: deep navy/void background (#0a0f1a), a vertical glowing divide line down the screen center with a pulsing orb at the midpoint, an animated starfield canvas split into blue stars on the left and purple on the right, radial gradients behind each half, full vignette overlay. Typography: DM Serif Display italic for headings and chapter titles, DM Sans for body text, Fira Code monospace for labels and metadata. Blue (#7aaeff) and purple (#a78bfa) accents with a shimmer gradient on the main title.
Form inputs: an age slider (18–85) with a large serif number display, fields for The Choice You Didn't Make, When You Faced This Choice, Your Actual Path, a Tone dropdown (Literary/Bittersweet/Hopeful/Sharp/Melancholy/Cinematic), a Detail Level dropdown (Rich & Novelistic/Quick Sketch/Decade by Decade), and an optional context textarea.
On submit, call the Claude API with a detailed system prompt that instructs it to return only raw JSON with: this_life_title, other_life_title, divergence (one evocative sentence), chapters array (period, title, body), and reflection (100-140 word closing that holds both timelines without choosing). Each tone has a specific prose instruction in the system prompt.
Render the result as: a split header showing Timeline A vs Timeline B labels with colored indicator dots, a divergence banner, chapter cards with period metadata, chapter title, and body text, a purple-accented closing reflection card. Include Copy, Rerun, and Reset buttons. Session history at the bottom logs past simulations by age and choice. Make it work in a single HTML file.What this does:
The Parallel Life Simulator takes your current age, one unchosen path, and the moment you faced it, then uses Claude to write a full novelistic alternate timeline narrated chapter by chapter from that divergence point to now. Each chapter covers a period of years with specific details, named moments, and real emotional texture. Six tone modes shape the prose from Ishiguro-style literary weight to cinematic scene-writing to sharp and unsentimental. A closing reflection holds both lives at once without choosing one over the other. Every simulation can be copied as a complete piece of writing, rerun for a different version, or saved to the session history alongside past simulations.
What this looks like:
WHERE WE STAND(based on today’s news)
✅ AI Can Now: Negotiate and close real-money deals between AI agents on behalf of human users (Anthropic's Project Deal hit $4,000 in 186 trades)
❌ Still Can't: Reliably diagnose patients with unclear symptoms (frontier models got it wrong over 80% of the time per JAMA Medicine)
✅ AI Can Now: Rebuild a live concert as a navigable 3D world from recorded video and audio (Coachella + DeepMind's Project Genie)
❌ Still Can't: Write a normal-sounding email without people sniffing out it's AI (hence Sinceerly, the plugin that adds typos back in)
RECOMMENDED LISTENING/READING/WATCHING
Al Pacino plays a director whose career is falling apart, so he creates a completely digital actress using software and passes her off as a real person. The film is a satire about celebrity, authenticity, and how easily audiences accept what they're told is real. It came out before deepfakes were a thing, before AI-generated faces, before virtual influencers. Watch it now and it feels less like a comedy and more like a document. The ending is dark in a way that didn't fully land in 2002 but lands very differently today.
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! Also free to check out our new site www.beginnersinai.org
-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.
