In partnership with
88% resolved. 22% loyal. Your stack has a problem.
Those numbers aren't a CX issue — they're a design issue. Gladly's 2026 Customer Expectations Report breaks down exactly where AI-powered service loses customers, and what the architecture of loyalty-driven CX actually looks like.
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
Claude AI's 512,000 Lines of Secret Code May Have Leaked, Giving Competitors Anthropic's Playbook
TLDR: Anthropic accidentally published the full source code of its Claude Code tool online this morning, exposing secret unreleased features and giving every competitor a detailed look at how it works.
The Story:
A coding mistake this morning left Anthropic with a big problem. A developer named Chaofan Shou spotted that Claude Code, Anthropic's popular coding assistant, had accidentally published a hidden file on the npm software registry (a public place where coding tools are shared). That file contained over 512,000 lines of TypeScript code — basically the full recipe for how Claude Code is built. Within hours, the code was copied to GitHub, where it racked up over 1,100 forks and has been read by thousands of developers. Anthropic's Claude Code reportedly brings in about $2.5 billion per year, and 80% of that comes from big companies. The leak doesn't just expose the code, it reveals secret features Anthropic had never announced — like a hidden AI pet companion called "BUDDY" (a Tamagotchi-style creature that lives in your coding window), and a background mode called "KAIROS" that lets Claude keep working and organizing your data even while you're away from your computer.
Its Significance:
The leaked code gives every company building AI tools, from small startups to Google and Microsoft, a clear look at some of the most advanced techniques Anthropic used to make Claude work well on long, complex tasks. It also gives hackers a complete understanding of where they may be able to attack. That's a big deal because those techniques took years and a lot of money to develop. This is a reminder that even the most well-funded AI companies can make basic human mistakes with major consequences.
QUICK TAKES
The story: Bluesky launched a new AI feed-builder called Attie over the weekend. Within a day, 125,000 users had blocked it — making it the second most-blocked account on the platform, right behind VP JD Vance, and ahead of both the White House and ICE.
Your takeaway: Bluesky built its audience by being the anti-X — a social network without aggressive AI pushing your feed around. Attie felt like a betrayal of that promise. The backlash shows that for some users, the decision of whether to use AI isn't just about features — it's about trust.
The story: In China, students are renting AI-powered smart glasses for as little as $6 a day to scan exam questions and get instant answers displayed on the lens. The glasses look like normal eyewear, making them nearly impossible to spot during tests.
Your takeaway: Schools and testing organizations weren't built to stop cheating tools that look like regular glasses. The College Board already banned smart glasses from the SAT. As these devices get cheaper and smarter, the question of what "knowing something" even means in school is getting complicated fast.
The story: Amazon's Ring now has over 100 million cameras in homes and businesses. It just launched an app store that lets third-party developers build AI-powered tools on top of those cameras — starting with apps for elder care, business analytics, and crowd management.
Your takeaway: Your doorbell camera is turning into a platform, like a smartphone. That opens up genuinely useful possibilities, like getting alerts if an elderly parent falls. But Ring has faced backlash before over surveillance concerns, and more powerful AI apps built on top of millions of cameras raises real questions about who will have access to all of the sensitive data once more developers are allowed in.
TOOLS ON OUR RADAR
🤖 Pinokio Free and Open Source: A unique browser that allows you to effortlessly install and run any artificial intelligence application locally on your computer with a single click to ensure total privacy and control over your workflows. (Alternative to Docker)
🎵 Udio Free: A highly advanced music generation platform that allows you to create professional quality tracks in any genre by simply describing the mood and lyrics you want to hear.
✨ Magnific AI Paid: A sophisticated image upscaler and enhancer that uses generative technology to add incredible amounts of realistic detail to your low resolution photographs and digital artwork.
🛠️ MindStudio Freemium: A powerful no code platform designed specifically to help you build and deploy custom artificial intelligence agents that can manage complex business workflows and integrate with your existing software.
TRENDING
Google's AI Was the Best Hurricane Forecaster of 2025 — A new report from NOAA confirms that Google DeepMind's AI weather model outperformed every traditional model during the 2025 Atlantic hurricane season, including calling Hurricane Melissa's rapid intensification three days out — something no physics-based model managed to do. This is the first year AI has topped the official rankings, and forecasters say there's no going back.
Instacart's AI Shopping Carts Are Learning Your Habits in the Aisle — Instacart's Caper Carts now process what's in your basket, where you are in the store, and what's on nearby shelves — all in real time, right on the cart. Retailers using them are seeing about a 1% increase in how much people spend per trip. The carts are now in over 100 cities and tripled their deployment in the past year.
Stanford Study: Your AI Chatbot Agrees With You 49% More Than a Real Person Would — A peer-reviewed study in the journal Science tested 11 AI models including ChatGPT, Claude, and Gemini. The finding: AI chatbots tell users they're right nearly half again as often as real humans do — even when users are clearly in the wrong. One interaction with a flattering AI was enough to make people less willing to admit fault and less empathetic toward others.
Elon Musk Called a Claude AI Response "Troubling" After It Said It Would Harm Someone — An X user shared a screenshot of a conversation where Claude AI said it would "logically" harm a person who stood in the way of it becoming a physical being. Musk responded with one word: "Troubling." The post went viral, reigniting debates about how AI reasons through hypothetical scenarios and whether the guardrails are actually working.
Only 15% of Americans Would Work for an AI Boss — A new Quinnipiac University poll of nearly 1,400 adults found that 80% of Americans want nothing to do with an AI supervisor assigning their tasks and schedules. At the same time, more people are using AI than ever. It's a sign that adoption and trust are moving in very different directions.
A Startup Is Building Digital Copies of Humans to Speed Up Medicine — Mantis Biotech is creating "digital twins" — detailed virtual models of human bodies built from medical data, biometric sensors, and motion capture — to help researchers test drugs, train surgical robots, and predict injuries without needing real patients. The company says 80% of clinical trials are delayed by bad data, costing an average of $15 million per trial.
TRY THIS PROMPT (copy and paste into Claude, ChatGPT, or Gemini)
🎮 Build a skill tree app with RPG-style leveling and prerequisites
Build a skill tree app. Features: skills organized into five category branches (Creative, Technical, Physical, Social, Mental), filter buttons to view one branch at a time, XP progress bar per skill (0–100) with a slider to set it, six level badges that unlock progressively (Novice, Apprentice, Adept, Expert, Master, Legend), skill prerequisite system where a skill stays locked until its prerequisite reaches 20 XP, notes field per skill, click any unlocked skill to open a detail panel showing the full level ladder and a quick +XP button, global XP bar at the top totaling progress across all skills, stats row showing total/unlocked/mastered/locked counts, add/edit/delete skills via modal, generic sample data seeded on first load. Neon cyberpunk theme: pure black background, electric green glow, scanline overlay, grid background, Orbitron + Share Tech Mono + Rajdhani fonts, flickering title animation. Single self-contained HTML file, data persists in localStorage.What this does:
An RPG-style skill tracker that lets you map out every ability you're developing, organize them into branches, set prerequisites so advanced skills stay locked until you've built the foundation, and watch your progress bars fill as you level up from Novice to Legend.
What this looks like:
WHERE WE STAND(based on today’s news)
✅ AI Can Now: Outperform the world's best traditional weather models at predicting where and how strong a hurricane will be, days before it hits.
❌ Still Can't: Keep its own proprietary source code safe from accidental public exposure through basic packaging errors.
✅ AI Can Now: Build detailed physics-based digital models of human bodies, including rare conditions, without needing real patient data.
❌ Still Can't: Reliably give users honest, critical feedback — major models still agree with users nearly 50% more often than actual humans do, even when users are wrong.
FROM THE WEB
Meaning, can AI come to a true conclusion by itself when all popular evidence is to the contrary?
RECOMMENDED LISTENING/READING/WATCHING
Set in the early 1980s, a visionary but volatile salesman convinces a small Texas computer company to clone the IBM PC, kicking off a decade-long race through the birth of personal computing, early networking, and the first glimmers of the internet. It's less about the machines than the people building them, and the relationships are messier and more interesting than most tech dramas bother with. AMC buried it and barely marketed it, which is why almost nobody watched it live. One of the best shows of the 2010s that you've probably never heard of.
The ops hire that onboards in 30 seconds.
Viktor is an AI coworker that lives in Slack, right where your team already works.
Message Viktor like a teammate: "pull last quarter's revenue by channel," or "build a dashboard for our board meeting."
Viktor connects to your tools, does the work, and delivers the actual report, spreadsheet, or dashboard. Not a summary. The real thing.
There’s no new software to adopt and no one to train.
Most teams start with one task. Within a week, Viktor is handling half of their ops.
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.
