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 published with the intention of making AI news and technology more accessible to everyone.

THE FRONT PAGE

What Happens If the AI Hype Dies? A History of AI Winters

TLDR: Peter Thiel just dumped all his Nvidia stock, comparing today's AI boom to the 1999 internet bubble—and AI has crashed before.

The Story: Peter Thiel completely exited his Nvidia position in Q3, selling over 537,000 shares that represented nearly 40% of his fund's portfolio. He's comparing the current AI boom to 1999, when investors priced in a future that would take 15-20 years to unfold—and he's not alone, with Jeff Bezos calling it an "industrial bubble" and Michael Burry taking massive short positions. But AI doesn't just have bubbles—it has winters. The first AI Winter (1974-1980) hit after early promises failed to deliver, triggered by the devastating Lighthill Report that criticized AI's "grandiose objectives" and led governments to slash funding. The second AI Winter (1987-1993) came when expert systems proved too complicated and expensive to maintain, causing corporate interest to collapse. Both times, the pattern was identical: massive hype → overpromising → underdelivering → funding freeze → years of research stagnation.

Its Significance: We might be setting up for a third AI Winter, but this time the stakes are dramatically higher. Previous winters affected mainly academia and niche tech companies—today, trillions in market cap hinge on AI delivering massive productivity gains soon, not in 2040. Unlike the 1970s when AI Winter was contained to research labs, today's AI is deployed at scale—ChatGPT has 100+ million users, companies have AI in production, and entire business models depend on continued advancement. If companies spending billions on AI infrastructure don't see ROI within 12-24 months (Goldman Sachs CEO's predicted timeline), we could see a market correction that makes the dot-com crash look modest. But there's another difference between then and now: unlike previous winters, today's AI actually works—it's just that the timeline to widespread economic impact might be a decade, not two years, meaning the technology won't disappear but rapid progress could freeze for years.

QUICK TAKES

Meta to Judge Employees on AI Skills Starting 2026

The story: Meta will start grading workers on their "AI-driven impact" in 2026, judging them on how well they use AI tools and build AI features. The company is also rolling out an AI assistant to help employees write their own performance reviews starting in December.

Your takeaway: Tech companies are making AI skills mandatory for career success, not optional, turning AI adoption into a job requirement.

Android Founder Launches Humanoid Robot Company in Tokyo

The story: Andy Rubin, who created Android, is reportedly starting a humanoid robotics company called Genki Robotics in Tokyo. He's setting up in Japan to tap into the country's deep pool of robotics engineers and university talent.

Your takeaway: Silicon Valley veterans are heading to Japan for robotics talent, signaling that the humanoid robot race is heating up globally.

Google Gemini and ChatGPT Now Let You Schedule AI Tasks

The story: Both Google Gemini and ChatGPT now let paid subscribers set up recurring AI tasks that run automatically, like daily weather reports or weekly meal suggestions. The feature costs $20 a month on both platforms and handles up to 10 scheduled tasks at once.

Your takeaway: AI chatbots are shifting from tools you ask questions to assistants that anticipate your needs and take action without being prompted.

MIT Professor Studies How AI and Human Intelligence Are Similar

The story: MIT researcher Phillip Isola found that different types of AI models—from language models to image recognition to audio processors—all seem to represent the world in similar ways as they get bigger and train on more data. His team calls this the Platonic Representation Hypothesis, suggesting AI models are converging toward a shared understanding of reality.

Your takeaway: As AI models grow more powerful, they may be discovering the same underlying patterns about how the world works, regardless of whether they process text, images, or sound.

Google Releases WeatherNext 2 AI Model

The story: Google's new WeatherNext 2 model generates weather forecasts 8 times faster than before and can predict hundreds of possible weather outcomes from a single starting point in under a minute. The model now powers weather forecasts in Google Search, Gemini, Pixel Weather, and Google Maps.

Your takeaway: AI weather models are getting fast enough to power everyday apps, bringing research-grade forecasting to billions of people in real time.

TOOLS ON OUR RADAR

• 🔨 Super [Free]: AI-powered search that connects all your company tools—Slack, Google Drive, Notion, Confluence—into one searchable hub with instant answers and custom assistants.

• 📐 InVideo AI [Free]: Turn text prompts into complete videos with AI-generated scripts, voiceovers, stock footage, and editing—no video skills required.

• 🔧 Airtop [Free]: Automate web browsing tasks using natural language commands—your AI agents can log in, scrape data, and complete complex workflows across any website.

• 🔨 Figma [Free]: Design and build complete products from first idea to functional app with AI-powered tools and team collaboration built-in.

TRENDING

Waymo Robotaxi Kills San Francisco Cat, Sparks Accountability Debate — The death of a neighborhood cat named Kit Kat by a Waymo robotaxi has reignited concerns about who's accountable when self-driving cars cause harm.

Cybersecurity Expert Questions Anthropic's Chinese Attack Report — A security researcher criticized Anthropic's report about a Chinese cyber attack on Claude for lacking technical details and proof that other organizations could use to protect themselves.

Essay Argues Only Three Types of AI Products Actually Work — A developer argues that despite billions in investment, only chatbots, code completion tools, and coding agents have proven successful as AI products.

AI Privacy Concerns Grow as Companies Track Everything — AI companies are collecting vast amounts of personal data through chatbots, smart devices, and tracking apps, with limited transparency about how that information is used or shared.

Google's DeepMind AI Beat Traditional Models in Hurricane Forecasting — Google's AI hurricane model outperformed traditional weather forecasting systems during the 2025 Atlantic hurricane season, giving forecasters more accurate predictions and earlier warnings.

TRY THIS PROMPT (copy and paste into ChatGPT, Grok, Perplexity, Gemini)

Analyze this recurring relationship conflict using psychoanalytic and attachment theory:

Situation: [Describe the recurring issue with your partner, family member, friend, or colleague—include what typically triggers it, what each person says/does, and how it usually ends]

Provide:

1. **Psychodynamic Analysis**: What defense mechanisms are both parties using (projection, denial, displacement, etc.)? What childhood patterns or family-of-origin dynamics might be replaying? What is each person's unconscious need or fear driving this conflict?

2. **Attachment Patterns**: Identify likely attachment styles at play (secure, anxious, avoidant, disorganized). How do these styles create a predictable dance of pursue-withdraw or other patterns? What does each person need to feel safe?

3. **Shadow Work**: What disowned parts of themselves might each person be seeing in the other? What would each need to acknowledge about themselves to break the pattern?

4. **Transference & Projection**: Who from the past might each person be unconsciously relating to through the other? What story from childhood is being reenacted?

5. **Actionable Scripts**: Provide specific "I feel/I need" statements that address the deeper emotional truth, not just surface complaints. Include repair attempts for after conflicts.

Draw on Bowlby, Winnicott, Jung, and modern relational psychoanalysis. Challenge my narrative about who's at fault.

What this does: Use this to understand recurring relationship conflicts at a deeper psychological level . Goes beneath surface-level communication advice to reveal the unconscious patterns, childhood wounds, and psychological defenses driving relationship conflicts—with concrete tools for healing and repair. *Do not take any advice from an AI as medical advice without consulting a real doctor.

WHERE WE STAND

✅ AI Can Now: Generate weather forecasts 8 times faster than before and predict hundreds of possible scenarios in under a minute on a single chip.

❌ Still Can't: Provide legal privacy protections for sensitive conversations people have with chatbots, leaving personal data vulnerable to subpoenas.

✅ AI Can Now: Schedule recurring tasks and run them automatically without supervision, acting as a proactive assistant rather than just answering questions.

❌ Still Can't: Match the accountability of human drivers when autonomous vehicles cause accidents, leaving victims without clear legal recourse.

✅ AI Can Now: Discover similar patterns about reality across different types of data—whether processing text, images, or sound—as models grow larger.

❌ Still Can't: Explain how it makes decisions or prove claims about security threats without transparent technical evidence.

FROM THE WEB

RECOMMENDED LISTENING/READING/WATCHING

Kai-Fu Lee has worked in tech in both the US and China, so he's got a unique view of the global AI race. This book opened my eyes to how different countries approach AI and what that means for the rest of us. It's less about the technical details and more about the bigger picture: who's winning, why it matters, and how AI is reshaping global power. Really eye-opening stuff.

CONTEST ANNOUNCEMENT (AI + Art + Lisbon)

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.