Sponsored by
Moda is the AI design agent with taste
Moda's viral launch hit 4.4 million views in two days. Tens of thousands of professionals signed up. Startups, agencies, forward-thinking brands and top firms are now using Moda to create brand-aligned slides, ad creative, reports, social carousels and more.
Most AI tools tend to create what we call "AI slop": repetitions of the same colors, layouts and fonts. And when you try to fix it, you get stuck in a loop of re-prompting.
Moda is different. Drop in your website URL, and Moda learns your brand from the ground up: your colors, your fonts, your visual language. Then it helps you generate pro-quality slides, docs, and marketing assets.
The best part? Every layer is fully editable on a real canvas, and exports to powerpoint, PDF and more.
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
Mastercard Just Let AI Spend Money on Its Own, Backed by 30+ Companies
LEAD STORY
TLDR: Mastercard launched a system that lets AI programs make and receive payments on their own, backed by 30+ companies including Coinbase, Ripple, and Stripe.
The Story:
On Wednesday, Mastercard rolled out Agent Pay for Machines, a payment system built for AI agents instead of people. The idea is simple. Software can now buy services, pay fees, and send money to other software without a person tapping a button. It works with cards, bank accounts, and stablecoins like Circle's USDC and Ripple's RLUSD. Mastercard says it's made for tiny, fast payments, some worth fractions of a cent, that happen quietly in the background. More than 30 companies signed on, among them Coinbase, OKX, Polygon, Stripe, Cloudflare, and the Solana Foundation. You can set spending limits and rules so the bots don't go wild.
Its Significance:
For years, paying for something online meant a person clicked "buy." That's starting to change. If AI agents can pay on their own, they can shop, book, and subscribe for you while you sleep. Handy, sure. But it also brings up a real worry: how much money do you trust a bot to spend before it checks with you? Mastercard is betting people will say yes, as long as they hold the controls.
QUICK TAKES
The story: Jalil Richardson was arrested in Florida after an AI system tagged him as an "85% match" to a car theft suspect, even though he was clocked in at work hundreds of miles away. He sat in jail over 50 days before his alibi cleared him, and by then he'd lost his job, his home, and custody of two of his kids.
Your takeaway: He's at least the 14th person that one sheriff's office has wrongly flagged this way. The software can sound confident and still be flat wrong, and the price gets paid by real people. Any program without human oversight will be vulnerable to this.
The story: A team of more than 20 researchers warned in Nature Communications that open-source AI is growing faster than governments can manage it. They say it can help tackle big problems like climate change and hunger, but it can also burn huge amounts of energy, widen the gap between rich and poor countries, and make fake content easier to spread.
Your takeaway: Their fix is four steps, including clear labels on AI-made images and video so people can tell what's real. Open tools are powerful and cheap to share, which is exactly why the team wants guardrails in place before 2030.
The story: Most AI models write one piece at a time, left to right. DiffusionGemma writes whole blocks at once and fixes mistakes as it goes, hitting more than 1,000 tokens a second on a top GPU (a token is a chunk of a word). It's free to download and can run on a home gaming card.
Your takeaway: Faster and local means tools that answer instantly without shipping your data to a cloud. The trade-off is quality. Google says its regular models still write better, so this one's tuned for speed, not polish.
TOOLS ON OUR RADAR
🚀 Spacedrive Free and Open Source: A futuristic file manager that connects all of your cloud storage accounts and hard drives together into one single beautiful folder interface.
📖 Wallabag Free and Open Source: A convenient read it later service that allows you to save web articles for offline reading while automatically removing distracting advertisements and popups for a cleaner viewing experience.
📊 Grist Freemium: A smart hybrid application that combines the simplicity of a spreadsheet with the power of a database allowing you to organize your business information visually.
📄 LibreOffice Free and Open Source: A powerful and complete office productivity suite that provides excellent alternatives to Microsoft Word Excel and PowerPoint without requiring costly software licenses.
TRENDING
Anthropic's CEO says AI is getting too powerful to wait on rules — Dario Amodei called for binding, FAA-style safety testing for the most powerful AI models, arguing that asking companies to simply be transparent isn't enough anymore.
Anthropic says AI is starting to help build the next AI — The company's research arm reports its engineers now ship about 8 times more code per quarter than a few years back, with Claude writing most of it, and warns this could push AI to improve itself faster than people can keep up.
AI is creating a hiring boom, just not for beginners — New research found 71% of AI job postings at big S&P 500 companies want experienced people, not entry-level hires, leaving new grads wondering how to get a foot in the door.
Google turns to Samsung for a piece of its next AI chip — With chip factories maxed out, Google may have Samsung make part of its 10th-generation TPU (code-named Icefish) rather than lean only on TSMC.
The next phase of AI in mental health care — The piece argues AI in behavioral health is moving past note-taking toward backing up a clinician's judgment, with humans staying in charge of the actual decisions.
AI designed tiny proteins that flip hard-to-reach drug targets on and off — Scientists used AI to build miniproteins that control GPCRs, the cell receptors behind about a third of all medicines. One matched a real drug in mice with fewer side effects.
AI is reshaping car design and factory floors — At an industry congress, experts said AI is speeding up car design and predicted humanoid robots could take over standing jobs on production lines by 2030.
Dario Amodei: "Policy on the AI Exponential" — In a new essay, Amodei laid out a plan to test frontier models across four risks (cyberattacks, bioweapons, loss of control, and AI running its own research), plus a separate framework for handling job losses.
TRY THIS PROMPT (copy and paste into Claude, ChatGPT, or Gemini)
🕯️ Name a memory that keeps replaying. See what it's actually pointing you toward right now.
Build a single-file HTML app with vanilla HTML/CSS/JS. The Recurring Memory Decoder — translate a memory that keeps replaying into what it's pointing at. Persist to localStorage key 'recurring_memory_v1'.
Aesthetic: deep midnight blue (#0d1428), candlelit atmosphere with golden radial gradient at top and muted purple at bottom, subtle noise texture overlay. Cormorant Garamond serif and EB Garamond italic for body, IBM Plex Mono for labels. Candle amber (#e8b870) primary with text-shadow glows, muted purple (#b48cdc) for the message card, soft sage (#80c89a) for action steps. Pulsing flame dot in header. Italic textarea, very low-key decode button with letter-spacing animation on hover.
Form: large textarea for the memory (with help text encouraging sensory details), when-it-surfaces dropdown (night / quiet moments / stressed / random / triggers / off-track), age-in-memory dropdown, recurrence-duration dropdown, feeling-it-leaves dropdown (warm but aching / restless / sad / confused / joy and grief / being shown something), current context textarea.
System instructions to the model: decode recurring memories. A memory that keeps replaying isn't random — it's pointing at something present-self needs to see. Read like a careful translator, not psychologize. Honor rather than analyze. Use their sensory details. Specific not generic. Return raw JSON: quiet_truth (one italic-worthy sentence of recognition), memory_type (evocative label from: anchor of safety / road not taken / unfinished business / template outgrown / preserved joy / warning ignored / moment of becoming / witness / goodbye unfinished / permission / last time something was simple), memory_description (2-3 sentences specific to their memory), the_message (2-3 sentences), whats_missing (2-3 sentences connecting to current context), echoes array (4 specific places this pattern echoes in present), what_to_do array (4 concrete moves), the_question (one specific italic-worthy question to sit with), closing (one gentle italic closing line).
Render: centered "quiet truth" card with large italic text bracketed by short gold lines. Two-column classify card with memory type label + description. Purple-bordered "message inside it" card. White "what's missing" card. Echoes card with ~ markers. Sage-bordered "what to do" numbered card. Centered "question to sit with" card with large italic text. Final closing line in italic with ~ marks. Section dividers with ◇ symbols and flanking dashed lines.What this does: Describe the memory the way it lives in your mind, when it tends to surface, your age in it, how long it's been recurring, the feeling it leaves, and where current you is right now. Get back a candlelit decoding: the quiet truth of what it's telling you, the memory's type classification (anchor of safety / road not taken / template you've outgrown / unfinished goodbye / etc.) with description, the message inside it, what present-you is missing, four echoes of its pattern in your current life, four concrete moves to honor what it's asking, the one question to sit with, and a closing line. Reverent rather than clinical. Saves to localStorage.
What this looks like:
WHERE WE STAND(based on today’s news)
✅ AI Can Now: Generate blocks of text in parallel at more than 1,000 tokens a second on a top GPU.
❌ Still Can't: Hit that speed and its best writing quality at the same time. The fast model trades polish for pace.
✅ AI Can Now: Design brand-new miniproteins that switch specific cell receptors on or off inside living cells.
❌ Still Can't: Reliably match a face from grainy security video to the right person, as a wrongful 50-day jail stay just showed.
RECOMMENDED LISTENING/READING/WATCHING
A novel that threads together a young Englishman in 1912, a writer living on a moon colony in 2401, and an investigator looking into strange anomalies across their stories. Mandel writes the most readable literary sci-fi alive, and this is her most playful book, with the AI and simulation questions sitting underneath everything without being announced. Bestseller, but routinely overlooked compared to her Station Eleven.
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.
What if AI found your next job overnight?
Stop wasting hours filling out repetitive applications. AIApply automatically discovers relevant jobs, optimizes your resume, generates personalized applications, and applies for you around the clock so you can focus on securing your next opportunity faster.