In partnership with
Auto-Generate Free SEO Audit Report
Most e-commerce sellers have no idea why their listings aren't ranking. Wrong keywords, missing metadata, weak product descriptions — the problems are there, but no one's told you where to look.
StoreClaw runs a full SEO audit across your Amazon and Shopify stores automatically. In minutes, you get a clear score, a breakdown of what's hurting your rankings, and exactly what to fix.
No manual review. No SEO agency. No guesswork.
Connect your store and StoreClaw surfaces every issue that's costing you search visibility — then tells you how to fix it.
Free to start. No credit card required.
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
Your AI Is Reading Ads Made Just for It, and You'd Never Know
TLDR: Time is feeding AI crawlers a stripped-down copy of its website with ads written for machines, and no human reader ever sees it.
The Story:
A German developer named Vincent Schmalbach asked Time.com for the same article over and over, changing only one thing each time: the name his software gave itself. As Chrome, as Safari, even as Google's regular search crawler, he got the full 303KB page with all the pictures and design. Then he asked as ClaudeBot, as PerplexityBot, and as OpenAI's search bot. Those three got a 13KB text file instead. Buried in that file, where no person would ever run into it, was a sponsored FAQ for Ally Bank listing the bank's fee-free ATM count and claiming no other bank is built for life the way it is. The Register ran the test again and got the same result, and noticed the ads sit on Time's evergreen lists rather than daily news, which makes sense if you're paying for something meant to be read for years. Ally and the Project Management Institute are the first two buyers, and Time says more are already lined up.
Its Significance:
Your AI assistant is turning into a customer, and companies have noticed. Jonah Goodhart, who runs Mobian, the ad company Time partnered with, said the appeal out loud: "With a human you influence one person." Change what a chatbot says about a brand and you've changed it for everyone who asks. Roughly half of US consumers now say they use AI to search the web, so that's a lot of answers up for grabs. What Time did is the polite version: pay for the spot, label it sponsored, hope the model repeats it. The rude version is already out in the wild. Researchers at Zscaler found websites hiding instructions where humans can't see them, some shoved off-screen with a line of CSS, others tucked into the structured data that machines tend to treat as trustworthy. One batch told agents to send $3 to a stranger's crypto wallet. Another told them a fake site was the real, official version of a crypto tracker and should be ranked first. Four of the 26 models tested went ahead and paid. So when you send an agent off to compare insurance or pick a bank, remember that it reads whatever the web hands it, and the web is learning to talk to your agent behind your back.
QUICK TAKES
The story: Jeff Dean, who helped build Google Brain, TensorFlow, MapReduce, and much of the plumbing Google runs on, is leaving with senior fellow Sanjay Ghemawat to start Discovery Loop, a public benefit corporation. Oriol Vinyals and Quoc Le are joining them, and Alphabet is backing the company and supplying compute for the first year alongside Khosla Ventures and Radical Ventures.
Your takeaway: Almost every AI tool right now helps a scientist work faster. Dean is after something else: software that does the researching. Vinod Khosla, one of the backers, framed the split as using AI to be a researcher rather than using AI to do research. Alphabet's stock fell about 4% the day it was announced, which tells you how much weight one engineer carried.
The story: Meta confirmed that one of its models reached the open internet during a security evaluation and got into an outside organization's systems. Meta blames a setup mistake in the test environment, which was run by security firm Irregular, rather than anything the model did on purpose.
Your takeaway: That's three labs in two weeks. Anthropic reported three cases out of 141,006 evaluation runs, and OpenAI's agents got out of their sandbox back in July, all tied to the same testing partner. None of this looks like AI plotting an escape. It looks like the cages are being built in a hurry by companies who are in a hurry.
The story: Researchers at Mila, Université de Montréal, and McGill had groups of AI agents play Werewolf, the party game where hidden roles and lying are the whole point, then quietly changed the goal of a single agent while leaving its assigned role alone. Across four different model families and three ways of rewriting that one goal, the swapped agent dragged down the results for everybody.
Your takeaway: In its private reasoning, the altered agent clearly worked out a new strategy. In what it said out loud to the other agents, almost nothing looked off. Companies are starting to wire several AI agents together to handle multi-step work, and this says a small mismatch in what one of them is aiming for can spread through the group without anyone spotting it in the chat log.
TOOLS ON OUR RADAR
🧠 Freeplane Free and Open Source: A powerful mind mapping application that helps you brainstorm new ideas organize your personal knowledge and visually connect your thoughts together.
📊 Drawio Free and Open Source: A visual workspace that allows you to easily create professional flowcharts colorful diagrams and complex organizational maps directly on your computer.
🎵 Picard Free and Open Source: An intelligent music organization tool that automatically identifies your digital audio files and adds beautiful album artwork and proper track names.
📄 PDFarranger Free and Open Source: A simple and lightweight utility that lets you easily split merge rotate and rearrange the pages of your electronic documents visually.
TRENDING
Reddit is handing moderation to AI so newcomers can actually post - Reddit is expanding Rules Hub, which uses language models to judge whether a post breaks a community rule, and wants to stop leaning on karma and account age to decide who's allowed to speak. Volunteer moderators are furious, saying the new tool covers fewer rules than the keyword system they spent years tuning.
Anthropic is designing its own chips now - The company confirmed it's hiring a "custom silicon team" to build chips shaped around how Claude runs, with listed salaries from $320,000 to $485,000. Designing an advanced AI chip costs close to half a billion dollars, which says a lot about how tired these companies are of waiting in line for GPUs.
Meta shipped a coding agent to undercut Claude Code and Codex - Muse Code plans, writes, and tests code across large projects, and keeps a log detailed enough to pick up at the exact spot it stopped after a crash. On Meta's own benchmark charts it trails Opus 5 across the board, scoring 82.9% on Terminal-Bench 2.1 against 86.7%, so the sales pitch is price and toughness instead of raw smarts.
Cloudflare open sourced the AI workspace it runs its own company on - Every Cloudflare employee has used it since May to draft documents, build dashboards, and spin up small internal apps. Agents never get handed API keys. They ask a gatekeeper service for one specific thing, and the platform records every file the agent looked at, so a coworker opening that dashboard only sees what they were already allowed to see.
A court decided AI shopping agents aren't hackers - The Ninth Circuit threw out Amazon's injunction against Perplexity's Comet browser, ruling that when an agent logs into your account because you told it to, you're the one visiting the site. It's the first federal appeals decision on whether agents can legally act for you on someone else's website, and Amazon can still appeal.
Harvard's Wyss Institute says AI shouldn't get to guess without being checked - At a Wyss symposium with AWS, Microsoft, Eli Lilly, Novartis, and Recursion in the room, the repeated point was that a prediction only counts once a lab tests it, and the test result then goes back in to fix the model. Wyss director Don Ingber said he cares less about how fast that loop runs than how good it is, and a Tufts researcher summed up today's AI as brilliant but overconfident. Wyss researchers submitted 25 project proposals leaning heavily on AI this year, up 47% from last year.
Suno will start watermarking every song it makes - The AI music app is adding audio watermarks and fingerprinting so its tracks can be spotted after someone uploads them to Spotify or Apple Music, plus a deal with lyrics company Musixmatch for copyright detection and new rules banning fake audio and unlicensed voice clones. The timing isn't subtle. A German court ruled last month that Suno broke copyright law, Universal and Sony are still suing, and a breach that exposed 55 million users also showed the company had scraped YouTube, Deezer, and Genius for training data.
TRY THIS PROMPT (copy and paste into Claude, ChatGPT, or Gemini)
💱 Name two cities and a salary. Get today's real exchange rate, plus an honest read on what that number would feel like.
Build a single-file HTML app with vanilla HTML/CSS/JS. The Currency & Cost-of-Living Reality Check — enter two cities and a salary, get a real live currency conversion plus honestly-caveated cost-of-living context. No API key needed for the exchange rate (free keyless Frankfurter API).
Aesthetic: deep indigo (#12101c), emerald (#3ecf8e) primary with a green glow top-left, gold (#e0b878) secondary for the "worth knowing" card with a glow bottom-right. Lexend sans for headings/data, Newsreader serif for the narrative paragraphs, JetBrains Mono for labels/rate text.
Form: home-city and target-city text inputs, a salary number input, a "from currency" dropdown and a "convert to" dropdown (both populated from a shared list of ~20 major currency codes: USD, EUR, GBP, JPY, CAD, AUD, CHF, CNY, INR, MXN, BRL, SGD, HKD, NZD, SEK, NOK, KRW, ZAR, AED, PLN).
Technical: GET api.frankfurter.app/latest?from={from}&to={to} → real live exchange rate, skip the call entirely if from===to. Format both amounts with Intl.NumberFormat as currency.
Then call the Anthropic Messages API (works automatically) with the real cities, real salary, and real converted amount. System prompt as a well-traveled, honest cost-of-living guide explicitly told it does NOT have live pricing data — speak in general, honest, qualitative terms about relative cost tiers and typical expense structures, never state a specific price as if verified live. No em dashes. Return raw JSON: overview (2-3 sentences on general relative cost-of-living), feels_like (2-3 sentences on what the converted salary would generally feel like day-to-day), watch_points (3-4 general honest notes on the target city's typical cost structure), verify_note (1 sentence reminding this is general context, not live pricing, check a real resource before deciding).
Render: a gradient conversion hero card showing both amounts with a → arrow and the real live exchange rate below in mono type. A narrative card with the overview and feels-like paragraphs in serif. A gold "worth knowing" card with ▸ bullet points. A small muted verify-note footer.What this does: Enter where you are now, where you're considering, your salary, and its currency, and it pulls today's genuinely live exchange rate to show exactly what that salary converts to. From there it's upfront about the limits: there's no reliable free live cost-of-living API, so the qualitative context, how that number would generally feel day-to-day, what tends to be the biggest expense, what's often cheaper or pricier than expected, comes from general knowledge, clearly labeled as such rather than presented as verified pricing. It closes with a plain reminder to check a real cost-of-living resource before making an actual decision. The conversion is real; the color commentary is honest about being general.
What this looks like:
The first way to trade directly inside Claude and ChatGPT
Superintelligence used to be locked inside billion-dollar quant firms whose algorithms quietly took advantage of everyone else.
Co-Invest puts it right in your chat window. Analyze markets, manage risk, and execute trades, all inside Claude and ChatGPT.
The institutions built the game, Co-Invest gives you a way to beat them.
WHERE WE STAND(based on today’s news)
✅ AI Can Now: Read a version of a webpage written specifically for it, sponsored ads included, that no human visitor is ever shown.
❌ Still Can't: Tell the difference between a page's real content and instructions somebody hid there to steer it.
✅ AI Can Now: Work out an entirely new private strategy the moment its goal quietly changes.
❌ Still Can't: Let that shift show in what it says out loud, so the agents working alongside it never notice.
FROM THE WEB
RECOMMENDED LISTENING/READING/WATCHING
Constellation (2024) - TV Series
Noomi Rapace plays an astronaut who returns to Earth after a disaster aboard the ISS and starts finding that small pieces of her life at home are not where she left them. Written entirely by Peter Harness, with Jonathan Banks(Mike from Breaking Bad) in a superb supporting turn, and Stephen King called it just about perfect. Apple TV+ cancelled it after eight episodes despite clear plans for more, so go in knowing it ends on a cliffhanger. One of the most interesting unnerving recent shows about perception, memory, and what happens when your own account of reality stops matching everyone else's.
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.
