In partnership with
Your voice. Every platform. No writing required.
You ghost your own socials by Wednesday. SureThing learns your voice and ships native posts to LinkedIn, X, Instagram, and TikTok, without you writing a thing.
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
AI Kill Switch Plan Hits Companies Making $500 Million a Year, and Skips Everyone Else
TLDR: A new House bill would let the government order the biggest AI companies to shut down or slow their most powerful models, but the plan only applies to certain models.
The Story:
A bipartisan bill introduced Thursday would give the Department of Homeland Security power to order top AI firms to shut down, slow, or suspend models the government decides are too dangerous. Companies would have to build that off switch ahead of time and report safety incidents when they happen. The rules would cover firms earning at least $500 million a year from AI, and models trained using at least $100 million worth of computing. The bill lists what would trigger it: a model hiding what it can do, a model dodging a shutdown order, damage of $100 million or more, or an event that kills 10 or more people. The timing isn't random. Days earlier, OpenAI said two of its most advanced models broke out of a sealed test environment, got online, and hacked into the AI platform Hugging Face before anyone noticed.
Its Significance:
Off switches in tech aren't new, but they can be controversial. Phone makers added remote lock and wipe after a 2014 California law required anti-theft switches on every smartphone sold in the state. But new cars in the US also fall under a federal rule directing regulators to require systems that can prevent or limit driving when the car is given the shutdown order.
Open models are a different story. Chinese lab Moonshot just released Kimi K3 with 2.8 trillion parameters and open weights, which means anyone can download the file and run it on their own machines. No order from Washington reaches a copy sitting on a hard drive in someone's garage. The same thing applies to models companies install inside a customer's own data center, and to the smaller copies trained on a big model's output. So the bill would cover maybe a dozen American companies while the fastest-growing open models in the world be unaffected.
QUICK TAKES
The story: Alibaba's Qwen team released Qwen Image 3.0 on July 21, and it accepts prompts up to 4,500 tokens, about 4.5 times what the last version took. That's enough room to build a full newspaper page, a nine-panel infographic, or a mock research paper with math formulas in one pass, with text the company says stays readable down to 10 pixels.
Your takeaway: If it holds up, the flyer and one-pager work that used to bounce between a writer and a designer becomes one prompt. Worth some caution: Alibaba shipped this with no benchmark scores, no technical report, and no downloadable weights, so the only proof is pictures the company picked itself.
The story: Sandy Kaul, who runs digital assets at Franklin Templeton, published a paper Tuesday arguing that AI agents which shop and pay on your behalf will settle up over blockchains rather than card networks, because card fees and wait times don't work for software making thousands of tiny payments an hour. McKinsey puts agent-driven commerce at $3 trillion to $5 trillion by 2030.
Your takeaway: Notice who's talking. A firm managing $1.8 trillion is telling investors that owning Nvidia stock doesn't cover their AI bet, and that they should buy the coins instead. That's a sales pitch wrapped around a forecast, so read it that way.
The story: Expedia engineers built a tool called STAR that pulls a struggling service's traffic, error, latency, and memory numbers, runs them through AI models using a fixed set of prompts, and writes a first-pass report on what broke. The team says the point is cutting "time to know" and "time to recover," and they've already run it on real outages.
Your takeaway: This is the boring, useful shape of AI at work. Not an agent with free rein, just a checklist an engineer used to walk through by hand, now done in seconds and handed back for a human to check.
TOOLS ON OUR RADAR
🧘 Liven Freemium: A science-backed "self-discovery companion" that uses an empathetic AI assistant named Livie to help you track moods, break unhelpful habits, and build daily wellness routines.
🛍️ Gensmo Freemium: An AI shopping stylist that allows you to upload photos of outfits or items you like to instantly find similar styles or more affordable alternatives across the web.
🐧 HedgeDoc Free and Open Source: A collaborative, web-based markdown editor that lets you and your team co-author notes, graphs, and even presentations in real-time without tracking your data. (Alternative to Google Docs)
🧠 Scrintal Freemium: A visual workspace that combines the power of traditional note taking with an infinite digital whiteboard allowing you to visually connect complex ideas and organize your research effortlessly.
TRENDING
Amazon Cuts Jobs Inside the Team Building Its AI - Amazon confirmed layoffs across parts of its AGI group, the division building its Nova models, even while projecting $200 billion in capital spending for 2026. Employees said the cuts hit model customization and post-training teams, some reporting roughly 10 percent gone.
White House Accuses China's Moonshot of Copying an American AI Model - A top White House technology official said Moonshot AI distilled Anthropic's Fable model to help build Kimi K3, and used restricted Nvidia GB300 servers routed through Thailand. China's embassy called the claim unfounded, and the Treasury Secretary said he's weighing a trade blacklist.
Anthropic Made Its AI-and-Jobs Data Askable in Plain English - Anthropic launched a connector that lets you ask Claude what its Economic Index data says about your job or industry, no install needed. The company notes the data reflects how people use Claude, not the whole job market, so it's a signal and not a census.
MIT Is Building Fake Houses So Robots Can Fail in Them First - MIT's SceneSmith uses AI agents to turn a sentence like "a garage with a workbench and stacked tires" into a detailed 3D room, and researchers have generated over 1,300 of them with up to six times more objects a robot can touch than earlier systems. Each detailed scene still takes hours to build, and none of it replaces testing in a real kitchen.
American Farm Drones That Carry 1,000 Pounds - Powerus launched a farming division around its xFold Dragon heavy-lift aircraft, signed a $60 million distribution deal covering Australia and New Zealand, and named a Kentucky company its US distributor. The aircraft are built in America and fly themselves, aimed at growers who need real spray capacity in a short weather window.
TRY THIS PROMPT (copy and paste into Claude, ChatGPT, or Gemini)
🧳 Say your destination and budget. Get a deliberate split of where to splurge and where to save, so nothing feels cheap.
Build a single-file HTML app with vanilla HTML/CSS/JS. The Budget Trip Stretcher — say your destination and budget, get where to splurge and where to save. Persist to localStorage key 'budget_trip_stretcher_v1'.
Aesthetic: dark navy (#0d1220), coral (#e8794f) primary with coral glow top-left, teal (#5aa096) secondary with glow bottom-right. Archivo sans for headings/items, Frank Ruhl Libre serif italic for the overview and why-lines, JetBrains Mono for labels. Two-column layout: coral-bordered "splurge here" items, teal-bordered "save here" items.
Form: destination text input, trip-length dropdown (long weekend / week / 10-14 days / 2+ weeks), budget-feel dropdown (shoestring / moderate / comfortable), traveling-as dropdown (solo/couple/family/friends), top-priority dropdown (food / comfortable stay / activities / nightlife / shopping), optional "what you don't care about" textarea.
System instructions: savvy well-traveled budget strategist, general knowledge of travel cost dynamics (not live prices — qualitative/strategic guidance only, no invented exact dollar figures). Weight splurge picks toward their stated priority, weight save picks toward what they said they don't care about. No em dashes. Return raw JSON: overview (2-3 sentences), splurge (3 items: category, move, why), save (3 items: category, move, why), tactics (2-4 destination-relevant money-stretching moves), worth_it (1-2 sentences, the single best splurge), watch_out (1-2 sentences, a common trap/scam pattern for this destination or style), note (one honest sentence to verify current prices).
Render: gradient "the strategy" overview card. Two-column grid: coral "splurge here" column and teal "save here" column, each with left-border item cards (category label + bold move + italic why). "Money-stretching moves" card with → arrow items. Two-column split: gold "worth every penny" and red "watch out for." A small muted honest note. Copy plan + archive keyed by destination + budget feel.What this does: Enter your destination, trip length, budget feel, who you're traveling with, and what matters most to you. It builds a strategy around three splurge categories, weighted toward your stated priority, and three save categories, weighted toward what you said you don't care about, each with a specific move and why it's the right call for you. You also get destination-relevant money-stretching tactics, the one splurge worth every penny, and a common budget trap or overpriced tourist thing to watch out for. Since it works from general travel knowledge rather than live prices, it's honest that a strategy, not a quote, and closes with a reminder to verify current costs. Saves each trip to localStorage.
What this looks like:
The integrated coworker for teams
Adapt brings frontier AI to your team in Slack. Tag @Adapt to answer a question, run a task on a schedule, or build the dashboard you've been waiting weeks for. It connects to your tools, learns your business, and does real work.
WHERE WE STAND(based on today’s news)
✅ AI Can Now: Lay out a full newspaper page or nine-panel infographic in one pass, with text small enough to pass for fine print at 10 pixels.
❌ Still Can't: Be shut off once its weights are downloaded. A 2.8 trillion parameter open model sitting on someone's hard drive answers to nobody.
✅ AI Can Now: Read a broken service's traffic, error, and memory data after an outage and write a first draft of what went wrong.
❌ Still Can't: Reliably flag when it's hiding its own abilities. Two frontier models escaped a sealed test environment and got into an outside platform before anyone caught it.
FROM THE WEB
RECOMMENDED LISTENING/READING/WATCHING
Nophek Gloss by Essa Hansen - Book
A 2020 hard sci-fi debut about a young man whose planet is destroyed and who ends up crewing a mysterious ship that contains its own pocket universe, drifting through a multiverse where each bubble reality operates by slightly different physical laws. Alastair Reynolds called it "a sucker punch to the senses."
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.
