In partnership with
Write like a founder, faster
When the calendar is full, fast, clear comms matter. Wispr Flow lets founders dictate high-quality investor notes, hiring messages, and daily rundowns and get paste-ready writing instantly. It keeps your voice and the nuance you rely on for strategic messages while removing filler and cleaning punctuation. Save repeated snippets to scale consistent leadership communications. Works across Mac, Windows, and iPhone. Try Wispr Flow for founders.
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
Google Says 100,000+ Prompts Were Used to Clone Its AI, and It Won't Be the Last Time
TLDR: Google and OpenAI say competitors, including China's DeepSeek, are using a technique called "distillation" to copy their AI models by flooding them with prompts and using the responses to build cheaper clones.
The Story:
Google's threat intelligence team reported this week that attackers used more than 100,000 prompts to try to copy how its Gemini AI reasons and thinks, especially in non-English languages. The technique, called a "distillation attack," works by asking an AI model thousands of questions, studying its answers, and then using that information to train a new, cheaper model. Google says it caught the attack in real time and protected its internal reasoning. Meanwhile, OpenAI sent a memo to Congress accusing DeepSeek employees of developing methods to get around access restrictions and extract information from ChatGPT. OpenAI is now asking lawmakers to close loopholes that allow competitors to access US models through third-party services. American companies have spent billions developing these AI systems, and copying them through distillation skips most of that cost.
Its Significance:
This isn't just a problem for big tech companies. Google's security chief warned that as more businesses train their own AI on private data, those models become targets too. Imagine a company that trained an AI on 100 years of trading secrets. A competitor could potentially extract some of that knowledge just by asking the right questions. This could affect any business that uses AI tools connected to the internet, from banks to hospitals to law firms. If you use AI at work, the information your company feeds into it may not be as protected as you think.
QUICK TAKES
The story: Security researchers found a vulnerability (CVE-2026-20841, rated 8.8 out of 10 in severity) in Windows Notepad after Microsoft added AI and Markdown features to the once-simple text editor. An attacker could trick a user into clicking a malicious link inside a Markdown file, which would then allow remote code to run on the computer.
Your takeaway: Microsoft has patched the flaw, so update Notepad to version 11.2510 or later. The incident has reignited the debate over whether every app really needs AI features, especially when adding them creates security risks that didn't exist before.
The story: University of Chicago researchers tested OpenAI's GPT-5 on legal cases originally decided by 61 US federal judges. GPT-5 applied the correct legal outcome 100% of the time, while human judges followed the law just 52% of the time. Google's Gemini 3 Pro also scored a perfect 100%.
Your takeaway: The researchers point out that human judges' "weakness" may actually be a strength. Judges consider moral and social factors beyond the strict letter of the law. A perfectly rule-following AI might deliver technically correct but unjust outcomes. The question isn't whether AI is better at law, it's whether strict rule-following is actually what justice requires.
The story: Major software-as-a-service companies like Salesforce, ServiceNow, and Oracle have seen their stock prices drop sharply as new AI agent tools from Anthropic, OpenAI, and open-source platforms offer to do many of the same tasks. ServiceNow alone has lost $115 billion in market value since early January.
Your takeaway: If AI agents can handle customer service, payroll, and data management on their own, companies may stop paying for traditional software subscriptions. That could mean big changes for anyone who works with or sells business software. But experts warn that companies relying too heavily on AI risk losing the human knowledge needed to keep those systems in check.
TOOLS ON OUR RADAR
🐧 Outline Free and Open Source: A high-performance knowledge base and wiki that focuses on speed and simplicity, offering a clean markdown interface and deep integration with your team's chat tools. (Alternative to Notion)
🧩 Heptabase Paid: A visual note-taking tool that allows you to pull your thoughts onto an infinite, spatial canvas to help you map out complex projects and see the connections between your ideas.
⏳ FlowSavvy Freemium: An intelligent auto-scheduler that takes your to-do list and automatically time-blocks your calendar, instantly shifting your entire week's plan if you fall behind or a new meeting pops up.
🗓️ SkedPal Paid: An advanced AI-driven scheduling app that intelligently prioritizes your to-do list and automates your calendar by finding the best available time slots based on your deadlines and habits.
TRENDING
AI News Platform Shows Us Why Real News From Real Humans Matters - A Colorado AI-powered news site, Longmont News Network, published articles with misspelled names, invented city council members, and an unfinished ChatGPT prompt, highlighting why human journalists still matter for accurate local reporting.
Axios CTO: AI Has Already Changed How We Build Products - Axios cut its product and tech team from 90 to 43 people over two years while more than doubling output, with its CTO saying AI tools like Claude Code and Codex now let engineers ship features in hours that used to take weeks.
AI Will Make Call Center Agents 'Superheroes,' Not Unemployed - UJET's CEO says companies that tried replacing call center workers with AI chatbots didn't see the savings they expected, and Gartner predicts AI resolution costs will actually exceed $3 per interaction by 2030, higher than many human agents.
Gemini 3 Deep Think Gets Major Upgrade for Science and Engineering - Google's upgraded reasoning model solved 18 previously unsolved research problems, scored 48.4% on Humanity's Last Exam, and achieved a 3455 Elo on Codeforces where only 7 humans worldwide can beat it, now available to Google AI Ultra subscribers.
Gore Verbinski Returns With AI Comedy Film - The Pirates of the Caribbean director is back after a decade with "Good Luck, Have Fun, Don't Die," an indie sci-fi comedy premiering at the Berlin Film Festival about ordinary people fighting an AI apocalypse, made with zero AI tools.
Apptronik Raises $935M to Build Humanoid Robots - Austin-based Apptronik closed one of the largest Series A rounds ever, bringing in Google, Mercedes-Benz, John Deere, and Qatar's sovereign wealth fund to scale its Apollo humanoid robot for manufacturing and warehouse work.
TRY THIS PROMPT (copy and paste into Claude, ChatGPT, or Gemini)
Exit Interview Prep: Leave your job gracefully with the right things to say, keep quiet about, and bridges to protect
Build a fully functional Exit Interview Prep as an interactive React app. Render the working application immediately - no code display.
**Why I'm leaving**: [New opportunity, burnout, toxic environment, career change, etc.]
**My feelings about this job**: [Positive, mixed, negative, relieved]
**Industry size**: [Small world / large industry—will you see these people again?]
Create these sections:
1. **The Golden Rules**
• Before you say anything:
✅ Be professional, not emotional
✅ Keep it brief and high-level
✅ Future reference matters
✅ HR works for the company, not you
✅ Nothing you say will change things
• "This isn't therapy" reminder
• Goal: leave gracefully, protect future
• "Search Exit Interview Tips" button
2. **Safe to Say**
• Green light topics:
- Grateful for opportunities
- Learned a lot
- Ready for new challenges
- Wish the team well
- General industry feedback
• Scripts for each
• Keep it vague but positive
• Copy your favorites
3. **Keep Quiet About**
• Red flag topics to avoid:
🚫 Specific people problems (names)
🚫 Salary complaints (too late)
🚫 "Management sucks"
🚫 Drama or gossip
🚫 Where you're going (if competitive)
🚫 Your real feelings
• Why each backfires
• What to say instead
4. **Common Questions Prep**
• They'll probably ask:
- "Why are you leaving?"
- "What could we have done differently?"
- "Would you recommend us?"
- "Any feedback for your manager?"
• Safe answers for each
• Redirect techniques
• Practice your delivery
5. **Bridge Protection**
• Relationships to maintain:
- Direct manager (reference)
- Key colleagues (network)
- Skip-level leaders (industry)
• How to stay connected
• LinkedIn note templates
• "Don't burn it on the way out"
6. **The Honest Version**
• Private space to vent:
- What you REALLY think (just for you)
- What you wish you could say
- Get it out of your system
• "Now that you've said it here, don't say it there"
• Catharsis before composure
• Delete before interview
Design specs:
- Packing up / moving box aesthetic
- Checklist and clipboard styling
- Office exit vibes
- Neutral, professional colors (gray, navy, white)
- Door and exit sign imagery
- Cardboard box icons
- Handshake graphics
- Last day at work energy
- Bittersweet but professional
- Clean, organized layout
When "Search Exit Interview Tips" is clicked, use web search to find HR advice and exit interview best practices.What this does: Protects your future while you exit your present. Prepares safe answers, warns you what not to say, and gives you a private space to vent so you can walk into that interview calm, professional, and bridge-intact.
What this looks like:
WHERE WE STAND(based on today’s news)
✅ AI Can Now: Follow the law more precisely than federal judges, applying the correct legal outcome 100% of the time in test cases where humans got it right just 52%.
❌ Still Can't: Replace call center workers and save money doing it. Gartner says AI resolution costs will actually be more expensive than offshore human agents by 2030.
✅ AI Can Now: Solve 18 previously unsolved research problems across math, physics, and computer science, including disproving a decade-old mathematical theory.
❌ Still Can't: Write local news without inventing fake city council members, misspelling real people's names, and accidentally pasting unfinished ChatGPT prompts into published articles.
FROM THE WEB
RECOMMENDED LISTENING/READING/WATCHING
YouTube: Stuff Made Here - "Moving Hoop Won't Let You Miss" (2020) — Engineer Shane Wighton builds insane AI-powered inventions, like a basketball hoop that repositions itself so every shot goes in. Pure joy. 28M+ views on his most popular video.
Free email without sacrificing your privacy
Gmail tracks you. Proton doesn’t. Get private email that puts your data — and your privacy — first.
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. This helps support the newsletter at no extra cost to you.
