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

Amazon Just Backed a $1.45 Billion Startup Building Worlds for Robots to Practice In

LEAD STORY

TLDR: Amazon just put money into Odyssey, a startup that builds pretend worlds where robots can practice, as Amazon races to own the next stage of AI beyond chatbots.

The Story:
Odyssey raised $310 million in new funding, which values the company at $1.45 billion. Amazon and chipmaker AMD are two of the new backers. That's a fast turn for Odyssey, which took money from Nvidia just a few months ago and now runs on Amazon's own chips instead. The startup builds what's called a "world model." Picture a video game world that follows real physics, so a robot can try a task thousands of times inside a computer before it ever touches a real factory floor. Odyssey has about 55 workers, and its founders came from self-driving car companies like Cruise and Wayve.

Its Significance:
Amazon is moving quickly into "physical AI," the kind that runs robots and machines instead of just answering questions. It has backed other startups, started a robotics fellowship with Nvidia, and Jeff Bezos is funding his own lab on the side. Training real robots is slow, costly, and sometimes risky, so a safe practice world could speed all of that up. For shoppers and workers, this points to a near future where more of Amazon's warehouses, deliveries, and machines lean on robots that learned their jobs inside a simulation first.

1-on-1 Live Claude AI Crash Course Tutorial
1-on-1 Live Claude AI Crash Course Tutorial
A 1-hour beginner-friendly video call to get you comfortable with the Claude ecosystem — Claude Desktop, Claude Code, Cowork, Skills, Projects, Correct File set-ups, and Plugins. Real-world example...
$75.00 usd

QUICK TAKES

The story: ChatGPT's share of AI chatbot users fell to 46% by the end of May, down from above 50% in March, per Sensor Tower's State of AI 2026 report. Google's Gemini holds 28% and Anthropic's Claude reached 10%, with Claude's U.S. share nearly tripling from 5% in December to 14% in May.

Your takeaway: ChatGPT still has more than a billion monthly users, so it isn't fading. The lead is shrinking because where an AI lives now matters as much as how smart it is. Gemini rides along inside Android, Search, and Chrome, and that built-in spot is hard to beat.

The story: MIT researchers made a long-term memory framework that lets a robot build a detailed mental map of a large space and recall it later using plain language. Ask it to "go grab the part we started building last night" and it can find the spot, and it answers questions more accurately than older methods while still running fast enough for a moving robot.

Your takeaway: A robot that remembers where it put things, and where you put things, is a real step toward helpers that work beside people in factories, warehouses, or homes. The same idea could power smart glasses that help you find your keys or guide you through a building.

The story: A writer had what felt like a stubborn calf cramp for five days, and a chiropractor treated it as a muscle problem. He ran his own records through an AI health tool he'd built, it flagged a possible deep vein clot (DVT) and told him to get an ultrasound fast, and the scan found four clots in his leg.

Your takeaway: He's clear that the AI didn't replace his doctors, who did the real work of treating him. It just helped him ask the right question at the right time. A new study in Science points the same way: an AI model was more likely than physicians to include the correct diagnosis on its list.

TOOLS ON OUR RADAR

🧩 ComfyUI Free and Open Source: A powerful visual workspace that lets you create high quality images by connecting different artificial intelligence modules together using a simple drag and drop system.

💬 Chatbox Paid: A beautiful desktop application that allows you to manage and chat with multiple artificial intelligence models in one highly organized and secure local interface.

🔒 PrivateGPT Free: A document assistant that allows you to upload and chat with your own private files locally on your computer without any data ever leaving your device.

🎥 Screenpipe Paid: A local desktop application that safely records and indexes everything you see and hear on your computer so you can search through your own history using natural

TRENDING

Snap's $2,195 AR Glasses, Specs, Open for Preorder - Snap revealed its first consumer AR glasses, Specs, at AWE 2026, with preorders open and shipping this fall in the U.S., U.K., and France. They cost $2,195, run on a Snapdragon chip, work without a phone, and tap OpenAI and Gemini for on-the-go help.

Qualcomm's New Chip Runs AI Right Inside Your Glasses - Qualcomm's new Snapdragon Reality Elite chip can run large AI models on the device itself, no cloud needed, with 160% more AI power than the last version. XReal's Project Aura glasses will be the first to use it, shipping this fall.

Microsoft's Copilot Cowork Goes Live for Everyone - Microsoft opened Copilot Cowork to all Microsoft 365 Copilot customers worldwide. It's an AI helper that finishes long, multi-step office jobs on its own, like drafting documents and building spreadsheets, and it now charges by how much work it does instead of a flat monthly fee.

OpenAI Tests New Models by Replaying Real Chats - OpenAI shared a method called Deployment Simulation that previews how a new model will act by replaying old, anonymized conversations through it before launch. It studied about 1.3 million chats to catch bad behavior early, and the trick works because models can't tell a replay from the real thing.

Android 17 Lands With New Multitasking and More Gemini - Google released Android 17 on Pixel phones first, adding a "bubble bar" for switching apps and a 50/50 split mode for foldables. A Pixel Drop also brought new AI: Lyria 3 makes music, Gemini Omni edits video by chat, and the Pixel 10a gets better live translation.

A Team Earned Top Marks at a Marine Corps AI Hackathon - At the Marine Corps Logistics Command AI hackathon during Modern Day Marine 2026, teams had only a few days to turn AI ideas into working prototypes for real supply and maintenance problems. One agency team earned top marks for building a brand-new prototype from a rough concept.

Pinterest Launches an AI Shopping App Called Ask Pinterest - Pinterest started testing a separate app, Ask Pinterest, that lets you shop by chatting instead of searching, like asking it to plan a dinner party on a budget. It uses Pinterest's "Taste Graph" of your interests and is limited to the U.S. for now.

TRY THIS PROMPT (copy and paste into Claude, ChatGPT, or Gemini)

Pick a topic. A passage flashes by at your target speed, then you're quizzed. Train to read faster while remembering more.

Build a single-file HTML app with vanilla HTML/CSS/JS. The Speed Reading Recall Gym  RSVP timed reading followed by a comprehension quiz, tracking the WPM-vs-recall tradeoff. Persist bests to localStorage key 'speed_reading_gym_v1'.

Aesthetic: near-black (#08090d), electric cyan (#50dcff) primary with glow, green (#50dca0) for comprehension/correct, soft red for wrong. Inter sans 800 headings, Newsreader serif italic for prompts, JetBrains Mono labels. WPM range slider with glowing thumb and tick labels.

Form: topic input, length dropdown (short ~120 / medium ~200 / long ~300 words), difficulty dropdown, display-mode dropdown (one word / 3-word chunks), WPM slider 150-800.

One API call generates the reading material. System instructions: write a tight self-contained factual passage at the length/difficulty target, plus 5 comprehension questions answerable ONLY from the passage testing specific recall (details, numbers, sequences, cause-effect a fast reader might miss). Return raw JSON: title, passage (clean prose, no markdown), questions array (5 items: q + options (exactly 4) + correct index varied + explain).

Then everything is client-side:
- Reader: "I'm ready"  3-2-1 countdown  RSVP presentation. Compute ms/word from WPM (60000/WPM). Show one word at a time (or 3-word chunks) centered and large, with the optimal-recognition-point letter highlighted cyan in single mode. Longer words get slightly more time. Progress bar fills. Track actual elapsed time  actual WPM.
- Quiz: passage hidden, 5 multiple-choice questions, select one each, submit enabled when all answered. On grade: mark correct green / wrong red, reveal explanations.
- Results: big actual-WPM next to comprehension %, verdict based on recall (80% strong / 50% mixed / <50% too fast), and speed-recall tradeoff coaching (push faster or slow down). 
- Persist: sprints run, top WPM achieved at 80%+ recall, average recall. "Same topic faster" regenerates at +75 WPM.

Render: header with sprint counter, form, loading, reader stage with countdown + RSVP word + progress bar, quiz blocks with option buttons, results card with two big stats + verdict + green tradeoff card, personal-bests bar (3 stats).

What this does: Choose a topic, length, difficulty, display mode, and a target reading speed (150–800 WPM on a slider). It writes a fresh passage and 5 comprehension questions, then runs the passage at you word-by-word (or in 3-word chunks) using rapid serial visual presentation with the optimal-recognition-point letter highlighted. The passage disappears, and you take the quiz on specific details a fast reader might miss. You get your actual WPM next to your comprehension score, an honest verdict on whether speed outran recall, and the speed-recall tradeoff coaching telling you whether to push faster or slow down. Tracks your top WPM at 80%+ recall, sprints run, and average recall in localStorage.

What this looks like:

WHERE WE STAND(based on today’s news)

AI Can Now: Run a full AI model right on a pair of glasses, with no internet connection needed.

Still Can't: Keep a detailed memory of a large, changing space as fast as a person can, though MIT's new robot memory framework is closing in.

AI Can Now: Build a pretend, physics-accurate world where robots safely practice real tasks thousands of times.

Still Can't: Guarantee that a skill learned in simulation carries cleanly into the messy, unpredictable real world.

FROM THE WEB

RECOMMENDED LISTENING/READING/WATCHING

A 2016 literary sci-fi novel about a near-future couple, a woman recovering from a personal tragedy and her physicist husband who is obsessively building what he refuses to call a time machine, who slowly start to feel that the world around them is somehow off-kilter. Palmer writes one of the smartest novels of the last decade about what it actually feels like to live inside the predictive systems running modern dating, journalism, and government. NPR, GQ, and the Washington Post all picked it as a best book of 2016, and somehow almost nobody outside sci-fi circles has read it.

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 happens when the S&P moves 3% during your commute?

We are living in volatile times. While you cannot control the state of international affairs, you can position your portfolio accordingly.

Liquid is one of the fastest growing trading platforms, allowing users to trade stocks, commodities, FX, and more 24/7/365 from their phone and computer.

Reply

Avatar

or to participate

Keep Reading