In partnership with
Better prompts. Better AI output.
AI gets smarter when your input is complete. Wispr Flow helps you think out loud and capture full context by voice, then turns that speech into a clean, structured prompt you can paste into ChatGPT, Claude, or any assistant. No more chopping up thoughts into typed paragraphs. Preserve constraints, examples, edge cases, and tone by speaking them once. The result is faster iteration, more precise outputs, and less time re-prompting. Try Wispr Flow for AI or see a 30-second demo.
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
Anthropic Safety Chief Quits to Write Poetry
TLDR: Anthropic's head of AI safety resigned in a cryptic letter warning that "the world is in peril," just days after the company released Claude Opus 4.6 and started chasing a $350 billion valuation.
The Story:
Mrinank Sharma led Anthropic's safeguards research team, the people who make sure AI doesn't do harmful things. He quit Monday with a resignation letter that quoted poets and talked about "a whole series of interconnected crises unfolding in this very moment." He didn't name specifics, but said constant pressure meant "setting aside what mattered the most to him: the values." One of his last projects? Understanding "how AI assistants make us less human or distort our humanity." Sharma's the third safety researcher to leave Anthropic in the past week. Harsh Mehta and AI scientist Behnam Neyshabur both left to "start something new." The timing's notable: Anthropic just shipped its latest model and is in talks for funding that'd value the company at $350 billion.
Its Significance:
When safety people leave AI companies, they don't usually write cryptic resignation letters about the world being in peril. They also don't typically mention that their final work was about AI making humans less human. One tech commentator nailed it: "The people building the guardrails and the people building the revenue targets occupy the same org chart, but they optimize for different variables. When the pressure to scale wins enough internal battles, the safety people don't fight forever. They leave and write beautifully worded letters about integrity." Three safety researchers leaving in one week, right as Anthropic pursues massive funding? That's not a coincidence. It's a pattern we've seen before at OpenAI. The question is whether anyone will pay attention before the next model ships.
QUICK TAKES
The story:
MIT researchers built an AI tool that can identify eight distinct nerve fiber bundles in the brainstem using regular MRI scans. The brainstem controls breathing, heart rate, consciousness, and sleep, but it's been nearly impossible to image in detail until now. The tool, called BSBT, revealed specific damage patterns in patients with Parkinson's, MS, traumatic brain injury, and Alzheimer's. In one case, it tracked a coma patient's seven-month recovery by showing how displaced brainstem bundles gradually healed and moved back into place.
Your takeaway:
This is the kind of AI work that actually matters, giving doctors tools to see things they couldn't see before. The brainstem controls some of your body's most critical functions, and we've basically been flying blind when it comes to imaging it. Now we can track damage and healing in real-time. That means earlier diagnosis, better treatment plans, and maybe even predicting who'll recover from a coma.
The story:
OpenClaw is a new AI platform that runs on your computer and can access your files, passwords, emails, and calendars to do tasks like organizing documents, writing emails, and building apps. It's got over 3,000 community-built extensions already. Northeastern University cybersecurity professor Aanjhan Ranganathan called it "a privacy nightmare." His concern: you're letting an AI agent look at all your sensitive data, but you've got limited insight into how it's processing that information or where it's sending it.
Your takeaway:
The professor's advice? "Set up my own virtual machine, set up a separate laptop, new email account, new calendars without giving it any real access." That should tell you everything. OpenClaw can do things standard AI chatbots can't because it has deep access to your system. But that's also why it's dangerous. Unless you're running it in a completely isolated environment with fake data, you're trusting it with everything on your computer. Most people won't set up a virtual machine. They'll just install it and hope for the best.
The story:
The Wikipedia competitor has become too big to ignore, with ChatGPT's latest model (GPT-5.2) now citing Grokipedia across various questions. Grokipedia is Elon Musk's AI-enhanced encyclopedia that has seen rapid growth in articles added and user citations, from 700,000 when it debuted to over 6 million today. Unlike Wikipedia, it doesn't let humans write the entries, but they can request changes. Anthropic's Claude has also been using Grokipedia.
Your takeaway:
Two of AI's biggest players are now using Grokipedia as one of their sources. Whether you see that as a problem or a solution depends on your view of Wikipedia's current level of neutrality. Wikipedia co-founder Larry Sanger has become one of his own creation’s ardent critics over recent years, posing the question: if Wikipedia's neutrality was compromised by ideological bias, what's the best alternative?
TOOLS ON OUR RADAR
🎙️ *Wispr Flow Freemium: Next-generation voice dictation tool that allows you to talk at 3x the speed of typing, automatically formatting your speech into structured text in any application while removing filler words.
🐧 Zulip Free and Open Source: Threaded team chat application that combines the immediacy of real-time chat with the productivity of email-style threading to keep conversations organized. (Alternative to Slack and Discord)
✨* Manus AI Freemium: An autonomous "research-first" presentation agent that doesn't just fill templates; it browses the live web to find credible data and evidence before structuring it into a professional narrative deck.
🧠 Read AI Freemium: A unified "workspace memory" that connects to your meetings, emails, and messages to build a personal knowledge graph, automatically surfacing relevant context from past discussions as you work.
TRENDING
University of Oklahoma Builds AI That Accelerates Antibody Drug Production - Researchers developed a machine learning model that correctly selected higher-performing cell clones in 76.2% of trials and predicted production outcomes using only the first nine days of data instead of several weeks. The tech could help drug companies get therapies to market faster and lower costs for patients. The market for monoclonal antibody treatments is forecast to double by 2030.
AI Tax Software Causes Wealth Manager Stocks to Tumble - Charles Schwab and Raymond James dropped Tuesday after startup Altruist announced AI-powered tax software that can read 1040 filings, paystubs, and emails to generate personalized tax insights for wealth advisors. Bloomberg analyst Neil Sipes said the sell-off shows investor fears about "efficiencies being competed away, fee compression long-term and potential market-share shifts."
Ring's Super Bowl Ad Sparks Mass Privacy Concerns Over AI Pet Tracking - Ring introduced "Search Party" during the Super Bowl, an AI feature that can access cameras throughout your neighborhood to help find lost pets. The backlash was immediate. The core concern: if Ring cameras can track pets with AI, they can track people too. Ring says facial recognition already exists and requires user permission, but that didn't ease fears. Pet owner Noel Lopez: "I'll probably look into another company instead of Ring."
Elon Musk Says AI Must Move to Space Within 30 Months or Hit the "End" - Musk told the Dwarkesh Podcast that AI operations need to relocate to space within 36 months (maybe 30) to solve power limitations. "All of the United States currently uses only about half a terawatt of power on average. Imagine trying to build enough power plants to double that." His solution: solar panels in space are "effectively 10 times cheaper because you don't need batteries."
Czech Ice Dancers Use AI Music at Olympics — It Plagiarizes Bon Jovi - Kateřina Mrázková and Daniel Mrázek performed at the Milano Cortina 2026 Winter Olympics to AI-generated music that included lyrics directly from New Radicals' "You Get What You Give" and phrases that sound exactly like Bon Jovi's "Raise Your Hands." The AI vocalist even sounds like Bon Jovi. It doesn't break Olympic rules, but it's a stark reminder that AI music models sometimes produce straight-up plagiarism.
Amazon Plans Marketplace Where Publishers Can License Content to AI Companies - Amazon's been meeting with publishing executives about launching a content licensing marketplace, following Microsoft's Publisher Content Marketplace launch. The move comes as the AI industry faces lawsuits over copyrighted training data and publishers worry that AI summaries in Google search are "devastating" their traffic. Publishers view the marketplace as "a more sustainable business that will scale up revenue."
Study Finds AI Doesn't Reduce Work — It Intensifies It - UC Berkeley researchers monitored a 200-employee tech company for eight months and found AI creates "workload creep" where employees take on more tasks than sustainable. "You don't work less. You just work the same amount or even more." The finding challenges promises that AI will make workers' lives easier and suggests benefits may not flow to rank-and-file employees.
TRY THIS PROMPT (copy and paste into Claude, ChatGPT, or Gemini)
Pros/Cons Weighted Calculator: Make better decisions by weighing what actually matters, not just counting items
Build a fully functional Pros/Cons Weighted Calculator as an interactive React app. Render the working application immediately - no code display.
**Decision I'm making**: [What you're trying to decide]
**Options**: [Option A vs. Option B, or just evaluating one thing]
Create these sections:
1. **List Builder**
• Two columns:
✅ PROS (reasons to do it)
❌ CONS (reasons not to)
• Add items to each side
• "Add Pro" / "Add Con" buttons
• Drag to reorder by importance
• Delete items you reconsider
2. **Weight Assignment**
• Not all factors are equal:
- Rate each item 1-10 importance
- "Deal breaker" flag for must-haves
- "Nice to have" flag for minor ones
• Slider for each item
• "Search Decision Weighting Methods" button
• Total weighted score per side
3. **The Real Score**
• Weighted calculation:
- Pros total: X points
- Cons total: Y points
- Net score: +/- Z
• Visual scale tipping left or right
• "It's closer than you thought" or "It's obvious" verdict
• Breakdown of biggest factors
4. **Emotional vs. Logical Split**
• Tag each item:
🧠 Logical (practical, factual)
❤️ Emotional (feelings, fears, desires)
• See balance: "You're 70% emotional on this"
• Neither is wrong—just know which is driving
• Gut check alignment
5. **Missing Factors Check**
• Prompt for things you might forget:
- Long-term consequences?
- Impact on relationships?
- What would you regret NOT considering?
- Opportunity cost?
• Add any that apply
• "What are you avoiding thinking about?"
6. **Final Verdict**
• The recommendation:
✅ DO IT: Pros significantly outweigh
🤔 CLOSE CALL: Need more info or gut check
❌ DON'T: Cons win clearly
• Key factors that swung it
• "If you're still unsure, that's data too"
• Save decision for future reference
Design specs:
- Balance scale / justice aesthetic
- Weighted scale tipping animation
- Clean, analytical dashboard
- Gold and dark wood legal vibes
- Sliding weights and measures
- Stacked bar charts for comparison
- Scale of justice imagery
- Numbers and percentages prominent
- Satisfying tip when one side wins
- Professional decision-making energy
When "Search Decision Weighting Methods" is clicked, use web search to find decision matrix techniques and weighted scoring frameworks.What this does: Upgrades the basic pros/cons list into an actual decision tool. Weights factors by importance, separates emotional from logical reasoning, checks for missing considerations, and gives you a clear verdict based on what truly matters.
What this looks like:
WHERE WE STAND(based on today’s news)
✅ AI Can Now: Map the brainstem's nerve fiber bundles in live MRI scans to track diseases like Parkinson's, MS, and traumatic brain injury in real-time
❌ Still Can't: Fully edit its own entries yet. Grokipedia still requires feedback and humans in the loop
✅ AI Can Now: Generate music convincing enough for the Olympics, even if it's plagiarizing Bon Jovi and New Radicals
❌ Still Can't: Make workers' lives easier despite the promises. UC Berkeley study shows AI just intensifies work instead of reducing it
FROM THE WEB
Seedance 2.0 is not available to most people yet. So far this and Kling are now the top video generators on market.
RECOMMENDED LISTENING/READING/WATCHING
A word association game where you type clues to help an AI match words, with blocks falling Tetris-style. It's fun and illustrates how language models understand semantic relationships. Free in your browser.
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 product links may be affiliate or referral links. This helps support the newsletter at no extra cost to you.
