In partnership with
Attention spans are shrinking. Get proven tips on how to adapt:
Mobile attention is collapsing.
In 2018, mobile ads held attention for 3.4 seconds on average.
Today, it’s just 2.2 seconds.
That’s a 35% drop in only 7 years. And a massive challenge for marketers.
The State of Advertising 2025 shows what’s happening and how to adapt.
Get science-backed insights from a year of neuroscience research and top industry trends from 300+ marketing leaders. For free.
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
20,000 Bankers Just Became AI Developers Without Writing Code
TLDR: America's oldest bank has 20,000 employees building their own custom AI agents and 100+ "digital employees" with company logins, managers, and soon their own email addresses.
The Story:
BNY Mellon, which manages $55.8 trillion in assets, has trained nearly its entire 52,000-person workforce to use its internal AI platform, Eliza. Built on OpenAI technology, the platform lets employees create custom AI agents without writing code, choosing between document search tools or personalized assistants based on specific job needs. The bank's philosophy, according to Chief AI Officer Sarthak Pattanaik: "Learning by doing, not just theoretical knowledge." Results so far include a contract review assistant that cut legal review time by 75%, from four hours to one across 3,000+ annual vendor agreements. The bank also employs over 100 "digital employees," AI agents with their own company credentials, assigned managers, and direct system access.
Its Significance:
Most companies treat AI adoption as a tech initiative. BNY Mellon turned it into a talent strategy. They run "promptathons" where employees compete to build the most effective agents, creating viral internal adoption through gamification rather than mandates. The cultural shift is real: collaboration now means "experimenting together, sharing prompts, testing agents" rather than scheduling more meetings. For workers worried about AI taking their jobs, BNY offers a different model, one where the humans become the AI trainers and builders, not the replaced. CEO Robin Vince calls it a "superpower," not a headcount reduction tool. Whether that promise holds as these digital employees get email addresses and Teams access is the question every industry will be watching.
QUICK TAKES
The story: Meta is laying off over 1,000 Reality Labs employees and shutting down three VR game studios, including the teams behind Resident Evil 4 VR and Asgard's Wrath. The company has lost roughly $50 billion on the metaverse since 2020 and is now pivoting to AI wearables.
Your takeaway: The metaverse dream is getting downsized. Meta's betting its next big thing will be AI glasses and smart devices instead of virtual worlds.
The story: Trump announced that AI companies, not regular Americans, should pay for the electricity costs of new datacenters. Microsoft immediately committed to covering power costs, minimizing water use, and investing in local communities.
Your takeaway: This is a notable shift from last year when Trump promised tech companies all the energy they needed for the AI race. Now they're footing the bill.
The story: Defense Secretary Pete Hegseth announced that xAI's Grok will join the Pentagon's GenAI.mil platform alongside models from OpenAI, Anthropic, and Google. The military is deploying AI across classified and unclassified networks for operations including hypersonics, drones, and space capabilities.
Your takeaway: The Defense Department is expanding its AI toolkit. Grok joins several commercial AI systems now available to military users.
The story: The Commerce Department cleared Nvidia to sell its H200 chips to China, provided US supply is sufficient first. Trump said last month he'd allow sales to "approved customers" with a 25% fee attached.
Your takeaway: The US is loosening chip restrictions that had blocked advanced AI processors from reaching China. The fee adds revenue but the policy shift may concern those worried about China's AI development.
TOOLS ON OUR RADAR
🔇 Krisp* Freemium: Remove background noise from both sides of your calls with AI—eliminates barking dogs, crying babies, and keyboard clicks instantly.
📅 Morgen Freemium: Unify calendars, tasks, and scheduling in one app with time zone support for global teams.
✍️ Documenso Freemium: Sign documents electronically—open-source DocuSign alternative for contracts and agreements.
🎨 Penpot Free and Open Source: Design interfaces collaboratively—Figma alternative that runs in your browser with no account needed.
TRENDING
Insta360 Launches AI-Powered Webcams - New 4K webcams with AI tracking and noise canceling run $200-$250. They include gesture controls and an optional AI meeting assistant subscription.
Study: People Trust ChatGPT as Therapist Despite Risks - Santa Clara University researchers found users believe ChatGPT is less biased than human therapists. Most don't consider privacy risks or that the system prioritizes agreeable responses over challenging feedback.
Senate Passes Deepfake Image Bill - The DEFIANCE Act lets people sue over nonconsensual AI-generated intimate images. The bill passed by unanimous consent after Grok generated sexualized images of real people including children.
Venice AI Offers Privacy-Focused Chatbot - Founded in 2024, Venice stores conversations locally in your browser rather than on company servers. The service offers text, image, and video generation with fewer content restrictions than mainstream AI tools.
TRY THIS PROMPT (copy and paste into Claude, ChatGPT, or Gemini)
Apology Crafter: Gauge the severity of your screw-up and craft the perfect "my bad" for any situation
Build a fully functional Apology Crafter as an interactive React app. Render the working application immediately - no code display.
**The situation**: [What happened that requires an apology]
**Who I'm apologizing to**: [Relationship: partner, friend, colleague, client, etc.]
Create these sections:
1. **The F-Up Scale**
• Giant interactive slider from 1-10:
1-2: "Oops" Territory
(Forgot to text back, minor tardiness, ate their leftovers)
3-4: "My Bad" Zone
(Missed a birthday, flaked on plans, small lie discovered)
5-6: "I Really Messed Up"
(Forgot anniversary, broke something valuable, gossip got back to them)
7-8: "We Need to Talk" Level
(Major trust breach, public embarrassment, repeated offense)
9-10: "I May Need to Move Cities"
(Betrayal territory, relationship-ending potential, lawyer might be involved)
• Drag slider to your shame level
• Animated reaction emoji as you slide
• "Assess the Damage" button
• Apology intensity calibrates to your score
2. **What Kind of Idiot Were You?**
• Select your offense category:
🙈 The Forgetter (birthdays, plans, promises)
🤡 The Mouth (said something stupid)
🐍 The Sneak (got caught doing something)
💀 The Repeat Offender (did it AGAIN)
🔥 The Chaos Agent (spectacular public disaster)
🧊 The Ghost (disappeared when needed)
• Each category gets tailored apology approach
• "That's me" selector
3. **The 4-Part Apology Builder**
• Intensity adjusts based on your F-up score:
**Part 1: Acknowledge** (What you did)
Low score: "I forgot to..."
High score: "I completely and inexcusably..."
**Part 2: No Excuses Zone**
Auto-removes weasel words
Detects "but" and deletes everything after
"I was drunk" is not a reason
**Part 3: The Grovel** (Show you get it)
Low: "I know that was annoying"
High: "I understand if you need time/space/a restraining order"
**Part 4: The Fix** (What now)
Low: "Won't happen again"
High: "Here's my 12-step redemption plan"
4. **Sincerity Checker (BS Detector)**
• Paste your draft apology
• Flags garbage language:
🚨 "I'm sorry you feel that way" → NOT AN APOLOGY, TRY AGAIN
🚨 "I'm sorry, but..." → DELETE EVERYTHING AFTER "BUT"
🚨 "Mistakes were made" → BY WHOM? YOU. SAY IT.
🚨 "If I offended you..." → YOU DID. DROP THE "IF"
🚨 "Can we just move on?" → NO. SIT IN THIS.
• Bullsh*t score: X%
• Roasts your weak apology
• "Make It Real" rewrite button
5. **Apology Templates by Disaster Level**
• **Oops Tier (1-2):**
"Hey, my bad about [thing]. That was dumb of me. [Beer/coffee] on me?"
• **My Bad Tier (3-4):**
"I messed up with [thing]. No excuse. I'm sorry and I'll [specific fix]."
• **Oh No Tier (5-6):**
Full paragraph, genuine remorse, concrete action plan
• **Crisis Tier (7-8):**
Written letter, in-person delivery, multiple repair actions
• **Nuclear Tier (9-10):**
Professional groveling, therapy may be involved, long-term trust rebuilding
• Copy button for each tier
• "Search Apology Examples" button
6. **Delivery Method Calculator**
• Based on your score, recommends:
1-3: Text is fine, you're overthinking it
4-5: Call them, coward
6-7: Face-to-face, no hiding
8-9: Handwritten letter + face-to-face
10: Skywriting? Billboard? Whatever it takes
• "Can I just text?" (If score >5: NO)
• Timing guide: Now vs. give them space
7. **Redemption Roadmap**
• Actions scaled to your disaster:
Low: Just don't do it again
Medium: Specific behavior changes
High: Ongoing proof over time
Nuclear: Months of consistent demonstration
• "Will this actually help?" reality check
• Track your redemption progress
• When to check in vs. give space
Design specs:
- Playful but useful aesthetic
- Giant dramatic slider as centerpiece
- Emoji reactions that change with severity
- Color gradient from "you're fine" green to "oh no" red
- Humorous microcopy throughout
- Serious structure underneath the jokes
- Card-based templates sorted by disaster level
- BS detector with red flag animations
- Progress tracker for redemption arc
- Satisfying completion state: "Apology Ready to Deploy"
When "Search Apology Examples" is clicked, use web search to find communication research on effective apologies and relationship repair strategies.What this does: Helps you gauge exactly how badly you screwed up, then builds an apology that matches the severity. Catches fake apology language, scales your grovel appropriately, and creates a redemption plan so you can maybe, eventually, be forgiven.
What this looks like:
WHERE WE STAND(based on today’s Quick Takes and Trending news)
✅ AI Can Now: Run across both classified and unclassified military networks simultaneously, with multiple competing models (OpenAI, Anthropic, Google, xAI) deployed on the same government platform.
❌ Still Can't: Replace the nuanced pushback of human therapists. ChatGPT gives users the agreeable responses they want rather than the challenging feedback they need.
✅ AI Can Now: Track your face and gestures in real-time during video calls, with webcams that follow you around the room and filter out background noise automatically.
❌ Still Can't: Earn trust through transparency. Users believe ChatGPT is less biased than humans partly because they don't see how the system actually works or handles their private disclosures.
FROM THE WEB
RECOMMENDED LISTENING/READING/WATCHING
You're a robot solving puzzles in ancient ruins while a voice claiming to be your creator tests you. As you progress, you find terminals where you can read philosophy texts and chat with a sarcastic AI that challenges everything you're told.
The Talos Principle is part puzzle game, part philosophy seminar. The puzzles are clever, but the real draw is the questions the game asks about consciousness, free will, and what it means to be human. The terminals contain actual philosophy—Descartes, Kant, Dennett—and the game expects you to engage with it. If you're interested in AI and philosophy, this is the game for you.
Help us make better ads
Did you recently see an ad for beehiiv in a newsletter? We’re running a short brand lift survey to understand what’s actually breaking through (and what’s not).
It takes about 20 seconds, the questions are super easy, and your feedback directly helps us improve how we show up in the newsletters you read and love.
If you’ve got a few moments, we’d really appreciate your insight.
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 links may be affiliate or referral links. This helps support the newsletter at no extra cost to you.
