In partnership with
Attio is the AI CRM for high-growth teams.
Connect your email, calls, product data and more, and Attio instantly builds your CRM with enriched data and complete context. Whether you’re running product-led growth or enterprise sales, Attio adapts to your unique GTM motion.
Then Ask Attio to plan your next move.
Run deep web research on prospects. Update your pipeline as you work. Find customers and draft outreach emails. Powered by Universal Context, Attio's intelligence layer, Attio searches, updates, and creates across your data to accelerate your workflow.
Ask more from your CRM.
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 Wrote Half of LinkedIn Posts in 2025, Now It's Being Punished
TLDR: LinkedIn's new AI system now spots AI-written posts and cuts their reach by about 30%, and other platforms like Pinterest and Instagram are following the same path.
The Story:
LinkedIn is done rewarding AI-generated posts. The platform's new ranking system, called 360Brew, can spot machine-written content and quietly cuts how many people see it. According to Engadget's coverage, posts flagged as likely AI-written get about 30% less reach and 55% less engagement than human-written ones, based on an Originality.AI study of hundreds of thousands of posts. More than half of long LinkedIn posts in 2025 were probably written by AI, so the floor of the feed is now machine-written and the platform is sorting it out. The shift isn't just on LinkedIn. Pinterest added a setting that lets users see less AI content after readers complained the feed was full of slop. Instagram chief Adam Mosseri said in late 2025 that AI content will likely overtake real photos, and the platform is rethinking how it labels what's real.
Its Significance:
If you use ChatGPT or Claude to draft posts and paste them straight in, your reach is probably dropping right now. Mid-level managers, sales reps, recruiters, and consultants who built audiences on LinkedIn are seeing posts that used to get thousands of views land in the hundreds. The fix isn't to stop using AI. It's to use it as a thinking partner, not a writer. Write your messy thought down first in your own words, then have AI help you sharpen it. The platforms are saying clearly that human voice and real expertise win now, and the AI shortcut is over.
QUICK TAKES
The story: Ken Griffin, founder of hedge fund Citadel and one of AI's loudest critics, just changed his mind on stage at Stanford. Work that used to take his teams of master's and PhD finance experts weeks or months is now done by AI agents in hours or days, and Griffin said he went home one Friday "fairly depressed" about what this means for society.
Your takeaway: When the head of a hedge fund that pays seven figures for elite quant talent says AI is replacing his top people, that's a real shift, not hype. Griffin's advice for workers: become a lifelong learner, because the people who keep growing their skills will be the ones who keep their jobs.
The story: Valar Labs got FDA Breakthrough Device Designation for Vesta Bladder Risk Stratify Dx, the first AI-powered digital pathology test for bladder cancer to earn that status. The test reads standard pathology slides that are already part of routine care and predicts which patients face aggressive disease versus slow-moving disease.
Your takeaway: Bladder cancer is one of the most common cancers worldwide, and doctors have long had to guess which patients need aggressive treatment. An AI that reads existing slides without new procedures means faster, cheaper answers for patients who would otherwise sit in a "gray zone" of uncertainty.
The story: Researchers at Carnegie Mellon built World2Rules, an AI that reads airport traffic data and historical crash reports to spot collision risks early. In direct testing, it was 23.6% more accurate than purely neural AI systems and 43.2% more accurate than older rule-based approaches.
Your takeaway: This kind of AI is built to work with human controllers, not replace them. After a near-miss at JFK in March between an Air Canada and EVA Air jet, tools like this one could give controllers an extra set of eyes that never gets tired or distracted.
TOOLS ON OUR RADAR
📺 Invidious Free and Open Source: A brilliant alternative interface allowing you to watch your favorite online videos completely free from annoying advertisements tracking algorithms and distracting recommended feeds.
💰 GnuCash Free and Open Source: A remarkably powerful financial accounting application that helps you balance your personal budget track your bank accounts and manage your investments safely offline.
✂️ Avidemux Free and Open Source: A straightforward and highly reliable video editing program perfectly designed for quickly cutting filtering and encoding your digital home movies without complex menus.
🔍 PhotoRec Free and Open Source: An incredibly powerful digital rescue utility that can miraculously recover deleted photographs videos and important documents from broken memory cards or accidentally formatted hard drives.
TRENDING
Ainos Wants to Teach AI How to Smell - Ainos is rolling out its AI Nose platform, which turns scent and chemical signals into machine-readable data. The company has a $2.1 million deal to deploy 1,400 units in semiconductor factories and is expanding into hospitals to spot infections, chemical leaks, and gas problems before humans notice.
Florida News Site Shut Down After Getting Caught Using AI Fake Reporters - The South Florida Standard published local news for months with fake AI-generated journalists, made-up headshots, and stolen content from real outlets. The site went offline after reporters from The Florida Trib started asking questions about its made-up editor-in-chief.
South Korea's LetinAR Raised $18.5M to Build the Lenses Inside AI Glasses - LetinAR makes the tiny optical modules that decide whether smart glasses feel like a clunky headset or normal eyewear. The startup, backed by LG Electronics, is gearing up for a 2027 IPO as Google, Samsung, and Apple all race to ship AI glasses.
Apple's New Siri Will Auto-Delete Your Chats - Apple's overhauled Siri, due at WWDC in June, will let users set chats to wipe after 30 days, one year, or never, similar to how Messages works. The standalone Siri app runs on Google's Gemini under the hood but on Apple's private servers, so Google can't train on your conversations.
ArXiv Will Ban Researchers for a Year if They Submit AI Slop - The biggest preprint research site is cracking down on papers with fake AI citations, leftover chatbot comments, and hallucinated content. A Lancet study found fake citations in biomedical papers jumped 12x since 2023, hitting one in every 277 papers in early 2026.
BBC: Foreign Networks Use AI Videos to Push UK Decline Story - The BBC traced dozens of Facebook and Instagram accounts running AI-generated "2050" videos showing British cities as dirty and chaotic. One account hit 20 million views, and London Mayor Sadiq Khan said some operators are profit-driven while others are backed by Russia, China, or US groups. Research shows people only spot AI fakes about 55% of the time, barely better than a coin flip.
US Regulators Use AI to Hunt Insider Trading on Polymarket - The Commodity Futures Trading Commission is feeding trade data into AI to spot suspicious bets on prediction markets like Polymarket and Kalshi. The agency is also chasing US users who hop onto blocked offshore platforms through VPNs.
New Nature Study: Governments Shape AI Chatbots by Shaping the Web - Researchers from Oregon, Purdue, UCSD, NYU, and Princeton tested chatbots in 37 languages and found that AI models give more favorable answers about countries with strong media control when asked in that country's language. Ask in Chinese, get a more pro-Beijing answer. Ask the same thing in English, get a different one.
TRY THIS PROMPT (copy and paste into Claude, ChatGPT, or Gemini)
✉️ Name who you want to reach. Get a cold email that sounds like a human wrote it, not a template.
Build a single-file HTML app with vanilla HTML/CSS/JS. Cold Email Lab — three versions, multiple subjects, follow-up included. Persist to localStorage key 'cold_email_lab_v1'.
Aesthetic: warm off-white (#f4f1ea), paper texture, black typography. Instrument Serif italic for headings, Inter for body, IBM Plex Mono for labels. Editorial feel.
Form: recipient, why them, the ask, what you bring, tone dropdown, relationship dropdown.
System instructions to the model: sound like one human writing to another. No "I hope this finds you well," no fake personalization, no marketing-speak. Short sentences. One clear ask. Return raw JSON: versions array (3 items: label, approach angle, body 80-130 words), why_works array (one sentence each), subjects array (4 items: tag like Curious/Direct/Specific/Pattern interrupt + text), followup (40-60 words, no whining).
Render: three-tab interface (A/B/C) each showing subject line, approach, body, and why-it-works italic block. Subject line options card below with tag badges. Follow-up card. Copy Active Version button. Archive in localStorage.What this does: Tell it who you're emailing, why them specifically, what you want, and what you bring. Get back three different angles (Version A/B/C), each with the approach explained and why it works. Plus four subject line options across different psychological hooks, and a follow-up message to send 5–7 days later. No "I hope this finds you well." No synergies. Saves to localStorage.
What this looks like:
Renewals stop being a fire drill.
Most churn blindsides the CSM in renewal week. Champion left. Usage dropped. NPS slid months ago.
A colleague in Slack watches the signals around the clock. Your CSMs catch every risk months before renewal.
11,000+ teams use Viktor daily. SOC 2 certified. Your data never trains models.
WHERE WE STAND(based on today’s news)
✅ AI Can Now: Read standard pathology slides to predict which bladder cancer patients face aggressive disease before treatment starts
❌ Still Can't: Write LinkedIn posts that pass the platform's new detector without losing 30% of normal reach if measures aren't taken to remove AI language patterns.
✅ AI Can Now: Spot suspicious trading patterns across millions of prediction market bets in real time
❌ Still Can't: Be used at face value in academic papers without risking a one-year ArXiv ban
FROM THE WEB
Interesting engineering here. Humans weren't optimized to move furniture around, but this robot was, with a fully rotating torso.
RECOMMENDED LISTENING/READING/WATCHING
An AI named Mother raises a single teenage girl in a sealed bunker after an extinction event, teaching her ethics and philosophy and grooming her to be the first of a new humanity. Then a wounded stranger arrives from outside and the story you've been watching starts to fall apart. Australian first-time director Grant Sputore made this on a small budget for Netflix and most people scrolled past it, which is a shame because the AI motive question at its core is well thought out and the movie keeps you guessing until the end.
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.
