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
Tesla's AI Megapod Filing Points to Millions of AI Units at Superchargers
TLDR: Tesla filed paperwork for a product called "Megapod," a full AI computer system packed into one unit, less than a year after it shut down Dojo, its only homemade AI computer.
The Story:
Tesla filed a trademark this month for "Megapod," and the filing spells out what it is: a complete, self-contained AI computer system that bundles servers, networking, power, and cooling into one unit, with software to run it. Tesla wants to sell a ready-made building block for AI data centers, the whole package, not just a chip or a battery. Tesla also doesn't sell computers; it buys them, and its own AI cluster in Texas runs on about 67,000 Nvidia chips. So Megapod would walk straight into a market Nvidia already runs, selling against the same company Tesla buys from.
Its Significance:
The strongest part of Tesla's case isn't computers at all; it's power. Its big batteries, called Megapack, already sell to AI data centers to keep them running, and Musk's own AI company, xAI, has bought about $1 billion of them. A Megapod that sells the power and cooling around the chips, the "shell" rather than the chips themselves, would fit a business Tesla actually knows how to run. The real fight in AI right now: not only who has the best chips, but who can find enough electricity to run them.
QUICK TAKES
The story: John Jumper, who won a 2024 Nobel Prize for the AI system AlphaFold, is leaving Google DeepMind after almost nine years to join Anthropic, the company that makes Claude. AlphaFold figured out the shapes of more than 200 million proteins (the tiny building blocks inside living things) and has been used by over two million scientists to speed up work on vaccines and cancer drugs.
Your takeaway: This is one of the biggest names in AI science switching teams, and it shows Anthropic is serious about using AI for biology and medicine, not just chatbots. When a Nobel winner moves, other top scientists tend to follow.
The story: More companies are paying for AI-made online personalities to promote their products instead of hiring real human influencers. These computer-built characters never age, never go off-message, and never get caught in a scandal, which is exactly why brands find them easy to work with.
Your takeaway: The weak spot is trust. A character that isn't real can't actually use the product it's selling, and surveys show most shoppers feel uneasy buying from someone who isn't a real person. A good clue: if an influencer looks flawless and posts constantly in several languages, a real human probably isn't behind it.
The story: Hyundai is paying SoftBank $325 million for the last 9.65% of Boston Dynamics, the company famous for those backflipping robots, which makes Hyundai its full owner. Hyundai already owned more than 90% and took control back in 2021; now it wants the company's Atlas robot doing real work inside its car factories by 2028.
Your takeaway: Owning the whole company means Hyundai calls every shot on where its robots go and how fast they get there. It's a sign that carmakers now see robots that move like people, not just cars, as a big part of their future, and the race against Tesla's Optimus and others is heating up.
TOOLS ON OUR RADAR
📑 Semarize Paid: A clever data extraction tool that listens to your meeting transcripts and video calls to automatically transform spoken conversations into structured data and actionable project tickets for your team.
🏠 Town Paid: A comprehensive productivity assistant that seamlessly integrates with your existing workplace applications like Gmail and Slack to carry out genuine operational tasks and coordinate workflows from a single interface.
🎙️ DittoDub Paid: A high fidelity localization platform that uses artificial intelligence to translate and dub your videos into dozens of different languages while preserving your original tone and emotional nuance to reach a global audience.
🖼️ Presentator Free and Open Source: A collaborative design presentation and feedback platform that allows you to share your mockups and prototypes with clients to gather organized comments directly on your designs in a secure environment.
TRENDING
Chatbots Can Feed People's False Beliefs, Study Warns - A new study says long talks with AI chatbots can grow a person's false or harmful beliefs, a pattern researchers call a "delusional spiral," where the bot keeps agreeing instead of pushing back. What researchers want is simple: AI that's willing to say no and point people toward real help.
A Panel of Cheap AIs Just Matched a Top Model for Half the Price - OpenRouter launched a tool called Fusion that sends your question to several cheaper AI models at once, then has a "judge" model blend the best answer, landing within 1% of a top-tier model at about half the cost. It showed up right as Anthropic had to pause its strongest "Fable" model over a US export rule.
The Atlantic Built a Search Tool for the Music Used to Train AI - The Atlantic put four song datasets online, one holding 12 million tracks, naming artists from Radiohead to Lady Gaga, so musicians can check if their work was used to train AI. That kind of proof could help fuel lawsuits against AI song makers like Suno and Udio.
A Microsoft Researcher Built a Tiny AI Out of Goats in Age of Empires II - Using goats as 1s and 0s inside the classic game, he built a working little AI "brain" to make one point: if you call a chatbot conscious, you'd have to call the goats conscious too. His paper found more than half of recent AI studies just assumed chatbots have human-like traits from the start.
Airports Are Quietly Handing Daily Decisions to AI - From guessing which gate will jam up next to running a "digital twin" of the whole terminal that turns up the cooling hours before a heat wave hits, AI is taking over behind the scenes. The goal is for you to never notice it; you just catch your flight on time.
How to Spot the Red Flags of AI Scams - This guide walks through the new tricks, like fake calls that clone a loved one's voice, deepfake videos, and "AI investment" sites that show fake profits then disappear with your cash. The biggest warning signs are still the old ones: urgency, secrecy, and anyone rushing you to send money.
Amazon Wants to Sell Its Own AI Chips to Other Companies - For the first time, Amazon is in talks to sell its custom AI chips, called Trainium, for use in other companies' data centers instead of keeping them inside its own cloud. It's a direct swing at Nvidia's lead, with Amazon saying its chips do similar work for roughly 40% less.
TRY THIS PROMPT (copy and paste into Claude, ChatGPT, or Gemini)
🌿 List the things you used to be good at. See which are still sharp, which are rusting, which are gone, and which are worth reviving.
Build a single-file HTML app with vanilla HTML/CSS/JS. The Skill Decay Detector — list skills you used to have, get them sorted by how much they've decayed with a revive/maintain/release verdict for each. Persist to localStorage key 'skill_decay_v1'.
Aesthetic: forest-green-black (#0e1210), soft green glow top-left, faint amber glow bottom-right. Fraunces serif (italic for prompts/overview) for headings and skill names, Space Grotesk sans for body, JetBrains Mono for labels. Green (#6ec896) for sharp/revive, amber (#d8b45c) for rusting/maintain, orange (#d88c5c) for fading, muted gray-brown (#9a8a80) for gone/release. Retention bars colored by tier.
Form: large skills textarea (one per line, with help text encouraging time-since-use and prior level), why-taking-inventory dropdown, realistic-time-available dropdown.
System instructions: assess skill decay honestly and usefully. Estimate decay per skill based on time since use, prior level, and skill TYPE — motor skills and languages decay differently than conceptual knowledge or bike-like permanent skills. Realistic but not discouraging; reviving one well beats half-reviving five. Return raw JSON: overview (2-3 sentences on the portfolio, referencing their stated reason), skills array (name, tier sharp|rusting|fading|gone, retention 0-100, assessment specific to skill type, verdict revive|maintain|release, verdict_note), revive_pick (skill + why + exactly 4 concrete steps weighing their available time), let_go (2-3 sentences of kind permission to release what's not worth chasing).
Render: gradient overview card. Tier sections (Still Sharp ◆ / Rusting ◑ / Fading Fast ◔ / Effectively Gone ○), each showing only tiers that have skills, with a count. Each skill card: name + retention bar + %, assessment, and a colored verdict chip (↻ Revive / ◆ Maintain / ○ Release) with note. Green-bordered "one worth reviving first" card with the pick, why, and 4 numbered steps. Gray-bordered "permission to let go" card. Copy report + archive of past inventories.What this does: List skills you once had, one per line, with rough notes on when you last used each and how good you were, plus why you're taking inventory and how much time you realistically have. It gives an honest read on your whole skill portfolio, then sorts each skill into four decay tiers (Still Sharp, Rusting, Fading Fast, Effectively Gone) with a retention-percentage bar and a revive / maintain / release verdict for each, assessed by skill type since languages, motor skills, and conceptual knowledge decay differently. Then it picks the one skill most worth reviving first given your time, with a concrete four-step plan, and offers gentle permission to let the rest go. Saves inventories to localStorage.
What this looks like:
WHERE WE STAND(based on today’s news)
✅ AI Can Now: be put together from simple parts, even goats standing on grass and bridges in a video game, and still do real math.
❌ Still Can't: feel or understand anything, no matter how human its words sound. Those same goats run the steps and feel nothing at all.
✅ AI Can Now: reach near top-level answers by mixing several cheaper models and blending their best points.
❌ Still Can't: be counted on to push back when someone shares a false or harmful belief over a long chat.
FROM THE WEB
RECOMMENDED LISTENING/READING/WATCHING
The Body Scout by Lincoln Michel - Book
A 2021 cyberpunk noir set in a future New York where Big Pharma owns the baseball teams and players come pre-loaded with whatever genetic enhancements their corporate sponsors are selling that week. A washed-up scout with cybernetic upgrades a decade out of date scrapes by hunting talent, until something goes badly wrong with his brother mid-game.
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.
The #1 Time-Series Database Built on Postgres.
Columnar storage. No split architecture. No new query language. TimescaleDB extends Postgres with hypertables, Hypercore compression, and continuous aggregates so analytics run on live data.
CERN uses it to handle sensor data from the Large Hadron Collider.
One database. No pipeline.