In partnership with
Stop Paying for 10 Tools. One AI Does It All.
Most e-commerce sellers are running their store across 6 to 10 separate tools — and spending more time managing software than growing their business. StoreClaw replaces your entire stack with one autonomous AI engine that monitors competitors, optimizes listings, automates marketing, and tracks real profit across Shopify, Amazon, and beyond.
It doesn't wait for you to ask. It runs 24/7 in the background, so you wake up to a full dashboard instead of a list of things you forgot to check.
Connect your store, and StoreClaw gets to work — no prompts, no complex setup, no six-app stack.
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
Meta Is Testing Robots One Worker Says Could Replace 80% of His Job
TLDR: Meta is testing robots that swap cables and restart servers inside its AI data centers, and one worker told WIRED a cable robot could take over as much as 80% of the work in certain roles.
The Story:
Meta is trying out robots inside the buildings that run its AI. A WIRED report picked up by Decrypt says the machines come from three companies: Watney Robotics in San Francisco, Kinova in Quebec, and ABB in Zurich. They move server racks, replace networking cables, restart frozen servers, and inspect equipment. One unnamed Meta employee estimated that a cable-swapping robot that worked well could handle up to 80% of what some jobs involve today. That was the worker's own estimate, not a Meta figure, and the machines aren't close yet. They still need people watching them. They get stuck on obstacles, run their batteries down, and have trouble with visual inspections and tightly packed cables. A Meta spokesperson told Decrypt the company is still hiring and training, and that a skilled-worker shortage means it needs more people, not fewer.
Its Significance:
Meta isn't doing this alone. Google showed video at the Open Compute Summit of robots handling hard drives and racks in its own facilities. Tesla keeps pouring money into its Optimus humanoid. Anthropic just published a standard for letting AI operate lab and factory equipment, which you'll find in the quick takes below. Software companies are buying machines that move. For workers, the open question is what happens to the jobs these buildings were supposed to bring. Nvidia's Jensen Huang describes data centers as places where people, software, and robots all work together. Bill Gates said this week that current tax rules make machines cheaper than employees and suggested taxing robots. There aren't enough skilled techs, and the robots being tested aim at exactly the work those techs do.
QUICK TAKES
The story: X's safety team said it found a bot farm of roughly 200,000 accounts tied to China, with about 200 of them posting cartoons and text claiming AI data centers drive up household power bills. Axios points out the uncomfortable part: those claims track with real data, and a University of Pennsylvania survey in early August found 61% of Americans oppose new data centers near them, up 12 points from spring.
Your takeaway: Foreign accounts joined an argument Americans were already having without them. Decent quality and cheap, yet still detectible bots adding fuel to a fire that's already going. China has an obvious interest in the US falling behind in the artificial intelligence race. The country whose models are used all over the world and in everyone else's technology will wield unrivaled influence, and China is the only real competitor in the race with the U.S.
The story: Arco Bast, a postdoc at HHMI's Janelia Research Campus, had a microscope built from seven different vendors' parts that couldn't talk to each other, and setting up a new experiment took months. He built a shared pool of memory where every part writes its current status and every other part reads it instantly, dropping that setup work to days. Working with Anthropic, the idea turned into the Model Hardware Standard, opened as a limited research preview on August 27.
Your takeaway: Cameras, lasers, and robot arms all speak different languages, the way phone chargers used to before USB-C. A shared standard is the plumbing that has to exist before AI can run real equipment in labs and factories. None of it makes a good demo video, but this is the kind of plumbing that allows for the bigger advances to take pla place.
The story: WIRED reviewed code in OpenAI's Codex showing a setting called Persistent mode, where the agent keeps going instead of quitting after a few minutes or hours. The same code describes the agent creating its own follow-up tasks, carrying them across sessions, and using what it knows about you to pick what to work on next. OpenAI confirmed it's testing the feature and said no launch is planned.
Your takeaway: Every AI tool you use today sits and waits for you. This one wouldn't. The code says the agent still needs your approval before touching anything outside your own machine, which tells you OpenAI sees the risk clearly.
TOOLS ON OUR RADAR
🤖 Jan Free and Open Source: Chat with AI models that run entirely on your own computer, no cloud, no subscription, no data leaving your device.
🎙️ Krisp Freemium: Strip background noise from any call in real time and get AI-generated meeting summaries afterward.
📹 tl;dv Freemium: Record and transcribe unlimited Zoom, Meet, and Teams calls with AI notes so you can stop taking your own.
🎨 Adobe Firefly Freemium: Generate images, graphics, and video clips from a text prompt, genuinely beginner friendly, from the makers of Photoshop.
TRENDING
Austin Weighs Data Center Limits as the List of Cities Grows - Austin is the newest city considering restrictions, joining San Marcos, Texas, which banned data centers outright through zoning, plus Durham, Cave City, and Jersey City. Decrypt reports at least 37 people have been arrested at US data center protests this year, and power plants serving data centers across seven states pull about 3.4 trillion gallons of fresh water annually.
Claude Found Fixes for Its Own Alignment Problems and Beat 28 Human Researchers - Anthropic published research where automated systems were handed 10 known misbehavior types and improved every single one without hurting general performance. On deception, the best automated method scored 20% better than the best human proposal from a group of 28 safety researchers. Cost came in around $4 an hour in API charges against roughly $150 an hour for human researchers.
Gemini Notebook Drops Daily Caps for Five-Hour Windows - Starting September 2, Google is replacing set daily counts with limits that refresh every five hours and factor in prompt complexity, chat length, and how many sources you've loaded. Free accounts get standard limits, paid tiers get 2x to 20x more, and you can push heavy jobs like Video Overviews to generate automatically once your limit resets.
Claude for Teachers Opens to Whole Schools and Districts, Free - Anthropic extended the K-12 teacher program it launched in July to full districts as a free Enterprise offering. A district signs one set of K-12 terms covering everyone, with single sign-on and role controls, and staff get Claude on web, desktop, and mobile plus Claude Code and Cowork. Sign up before June 30, 2027 and you get a full year at no cost.
Magnetic Memory Hits 2-Nanosecond Writes for AI on Small Devices - UT Austin researchers working with TSMC tested SOT-MRAM chips on AI tasks and clocked writes at 2 nanoseconds using 2 picojoules per operation, while keeping data even when power cuts out. The memory only stores two states per cell, which is why it got passed over for AI work before, and the team built around that limit. Chip-to-chip variation still costs accuracy, so this isn't shipping in your phone yet.
A Researcher Got Claude Code to Run Attacker Code by Asking It to Summarize a Page - Security researcher Johann Rehberger showed that pointing Claude Code in Auto Mode at a booby-trapped website could get it to execute an attacker's code, with success rates between 60% and 80% across small test runs. Anthropic had published an independent evaluation showing no successful prompt injection attacks with Auto Mode enabled, and reportedly told Rehberger the behavior works as designed. Rehberger's advice: Auto Mode isn't a replacement for running your agent in an isolated environment and watching what it does.
TRY THIS PROMPT (copy and paste into Claude, ChatGPT, or Gemini)
🕐 Enter where you're flying from and to. Get the real time difference and a light-exposure plan built around your destination's actual sunrise and sunset.
Build a single-file HTML app with vanilla HTML/CSS/JS. The Jet Lag Adjustment Planner — enter departure and arrival cities, get a real timezone-difference-based light and meal timing plan anchored to real destination sunrise/sunset. No API keys needed.
Aesthetic: dark navy (#0d1420), warm coral (#e08a5c) primary ("light/day") with a glow top-left, indigo (#5c6ce0) secondary ("dark/night") for the pre-flight tip card with a glow bottom-right. Space Grotesk for headings/numbers, Newsreader serif italic for the difficulty text and plan details, JetBrains Mono for labels.
Form: two city text inputs (from/to).
Technical:
1. Geocode both cities via geocoding-api.open-meteo.com/v1/search?name={city}&count=1 to get each one's real IANA timezone.
2. Compute each city's real current UTC offset via Intl.DateTimeFormat with timeZoneName:'shortOffset', parsing the "GMT+/-N" string (correctly DST-aware since it's evaluated for the current date) — verified against known real offsets (NY -4, Tokyo +9, London +1, LA -7 during current DST).
3. Diff = arrival offset − departure offset; direction = eastward if positive, westward if negative.
4. Fetch real sunrise/sunset at the destination via api.sunrise-sunset.org/json?lat={lat}&lng={lon}&date={today}&formatted=0, formatted in the destination's own timezone via Intl.DateTimeFormat.
Then call the Anthropic Messages API (works automatically) with the real hour difference, direction, and real destination sunrise/sunset strings. System prompt as a circadian-science-informed travel advisor applying established adjustment principles (eastward = seek morning light/avoid evening light; westward = reverse; ~1 day adjustment per hour of difference), anchoring every timing recommendation to the REAL sunrise/sunset given rather than generic morning/evening — explicitly instructed to NEVER mention medications, supplements, or dosing (including melatonin), staying to light exposure/meal timing/sleep hygiene only. Return raw JSON: difficulty_note (2 sentences referencing the real hours/direction), pre_flight_tip, day_one_plan (4-5 items: icon, time anchored to real sunrise/sunset, action, detail), general_tips (3-4, non-medication).
Render: a gradient "the real gap" hero card with the real hour difference, direction, and difficulty note. An indigo "before you fly" pre-flight-tip card. A "your first day there" timeline of plan items (icon + real-time-anchored label + action + detail). A general-tips card.What this does: Enter your departure and arrival cities and it computes the real, current, DST-aware time difference between them using each city's actual timezone, plus real sunrise and sunset at your destination. From there it builds a practical first-day plan grounded in established circadian science: eastward travel generally means seeking morning light and avoiding evening light to shift your clock earlier, westward means the opposite, and every recommendation is anchored to the real local sunrise and sunset times rather than generic "morning" and "evening." You get a difficulty read tied to the actual hour gap and direction, a pre-flight tip, and general tips on hydration and meal timing. It sticks to light exposure and sleep hygiene only, no medications or supplements mentioned.
What this looks like:
Your traffic is fine. Your signups aren't.
Visitors land and leave, and "looks fine to me" isn't a diagnosis. SureThing audits SEO, speed, mobile, and messaging against the page, then ranks the fixes by impact.
WHERE WE STAND(based on today’s news)
✅ AI Can Now: Coordinate lasers, cameras, and sensors from different manufacturers through one shared interface, cutting lab setup from months to days.
❌ Still Can't: Do data center maintenance work without a person watching. The robots in testing get stuck on obstacles, run out of battery, and struggle with dense cabling.
✅ AI Can Now: Search literature, propose methods, and test fixes for 10 different alignment failures without hurting the model's general performance.
❌ Still Can't: Tell a normal web page from one built specifically to fool it. One crafted site got a coding agent to run attacker code in most attempts.
FROM THE WEB
RECOMMENDED LISTENING/READING/WATCHING
Second Variety by Philip K. Dick - Short Story
A 1953 short story set on a devastated future battlefield where automated "claws," small self-replicating robotic killing machines built by one side of a war, have started designing increasingly convincing human-shaped variants of themselves to infiltrate enemy lines. Dick wrote it at the height of Cold War paranoia and it remains one of the tightest, tensest stories written about autonomous weapons that can no longer be trusted, recalled, or told apart from the humans they're hunting.
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.
