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

ChatGPT Preparing for In-App Ad Delivery, According to Recent Reports.

TLDR: A leak from ChatGPT's Android beta reveals OpenAI is building an advertising system, marking a major shift for a company whose CEO once called ads a "last resort."

The Story:

Code discovered in ChatGPT's Android app version 1.2025.329 includes references to "search ad," "search ads carousel," and "bazaar content" — clear signs OpenAI is testing how to display sponsored content. The initial rollout appears limited to search-related queries, mirroring Google's model. No users have reported seeing ads yet, but the company has been hiring advertising veterans from Google, Meta, and X. ChatGPT now serves roughly 800 million weekly users, up from 100 million in November 2023, processing an estimated 2.5 billion prompts daily. That scale makes it one of the largest untapped advertising platforms on the internet.

Its Significance:

Sam Altman's position on advertising has shifted considerably. He once called ads a "momentary industry" and said "ads-plus-AI is sort of uniquely unsettling to me." More recently, he's admitted Instagram ads changed his mind: "I love Instagram ads, they've added value to me, I found stuff I never would've found." The financial pressure is real — OpenAI spends over $8.5 billion annually while generating roughly $4 billion in revenue, and internal documents project $1 billion from "free user monetization" by 2026.

Altman insists ChatGPT won't let advertisers influence rankings, calling it "catastrophic" if the platform took money to promote worse recommendations. But every major platform has made similar promises before drifting toward advertiser priorities. The difference this time? ChatGPT collects something Google never could: extended conversations revealing users' jobs, problems, and aspirations. That's ad targeting at a depth no search engine can match. Bank of America analysts have already warned this could trigger budget shifts away from Google Search if the model gains traction. If this happens, expect all AI companies to follow OpenAI’s lead.

QUICK TAKES

NYC Nurses Push Back on Hospital AI Rollouts

The story: New York nurses are criticizing city hospitals for rolling out AI tools without consulting frontline staff. At a recent State of Nursing Committee meeting, nursing leaders described being surprised by new AI systems appearing on their floors, including devices hooked to ICU patients without warning, training, or protocols. Mount Sinai invested millions in an AI assistant called Sofiya for their cardiac catheterization lab, though nurses say they still have to double-check its work.

Your takeaway: This growing tension between hospital administrators adopting AI and nurses demanding input on patient care decisions could shape how healthcare AI gets deployed nationwide.

Google CEO Calls for National AI Regulation to Compete With China

The story: Sundar Pichai warned that over 1,000 AI bills moving through state legislatures could create a confusing patchwork of rules that weakens U.S. competitiveness against China. Speaking on Fox News Sunday, he called for federal-level regulation that balances innovation with guardrails, and urged governments to develop international frameworks so AI isn't weaponized between countries.

Your takeaway: The push for unified national AI rules is gaining momentum as tech leaders worry that state-by-state regulation could slow innovation while global competitors move faster.

MIT Releases Open-Source AI for Drug Discovery

The story: MIT researchers released BoltzGen, a generative AI model that can design novel protein binders for hard-to-treat diseases. Unlike existing tools that only work on "easy" targets similar to their training data, BoltzGen was tested across eight wetlabs on 26 targets—including previously "undruggable" ones. The open-source release drew over 300 attendees to its debut seminar and raised questions about whether paid binder-design services can compete when free alternatives keep improving.

Your takeaway: Open-source drug discovery AI could speed up treatment development for difficult diseases, while also disrupting the business models of biotech companies offering similar services.

Bipartisan Bill Targets AI Fraud With Steep Penalties

The story: Reps. Ted Lieu (D-CA) and Neal Dunn (R-FL) introduced the AI Fraud Deterrence Act, which would raise penalties for AI-assisted fraud to up to $2 million in fines and 30 years in prison for bank fraud. The bill also criminalizes using AI to impersonate federal officials—a response to recent incidents where scammers used voice cloning to mimic White House Chief of Staff Susie Wiles and Secretary of State Marco Rubio.

Your takeaway: As AI makes scams cheaper and more convincing, lawmakers are racing to update fraud laws that weren't written with deepfakes in mind.

Cursor Raises $2.3 Billion in Largest AI Startup Round of November

The story: Anysphere, the company behind AI coding tool Cursor, closed a $2.3 billion Series D that values it at $29.3 billion. It's the company's second major raise this year. Backers include Accel, Thrive Capital, Andreessen Horowitz, Nvidia, and Google. The deal is part of a broader surge in AI coding tools—49 U.S. AI startups have now raised $100 million or more in 2025.

Your takeaway: AI-assisted coding has become one of the hottest categories in tech, with investors betting that tools like Cursor will fundamentally change how software gets built.

TOOLS ON OUR RADAR

TRENDING

Researchers Bypass AI Safety With Poetry – A new study found that phrasing harmful requests as poems bypassed AI safety filters 62% of the time across 25 major models, including systems from OpenAI, Anthropic, and Google.

Jenna Ortega Calls AI "Terrifying" for Filmmaking – At the Marrakech Film Festival, the Wednesday star said AI feels like "Pandora's box" and hopes audiences will eventually reject AI-generated content as "mental junk food."

James Cameron Distinguishes His Work From Generative AI – While promoting Avatar: Fire and Ash, Cameron explained that performance capture celebrates actors' work, calling the idea of AI-generated performances from text prompts "the opposite" of what his films do.

Karpathy Explores the Difference Between Human and AI Minds – In a new essay, the former OpenAI researcher argues that LLMs are shaped by commercial pressures like upvotes and task rewards, while animal brains evolved for survival in adversarial environments.

AI Agent Learns to Use CAD Software – MIT engineers trained an AI on 41,000 videos to operate CAD programs click-by-click, turning 2D sketches into 3D models and potentially making design software more accessible.

TRY THIS PROMPT (copy and paste into Claude or Gemini(click on build first here))

What Should I Watch Tonight?: End the endless scrolling with an interactive recommendation engine that actually gets your taste

What this does: Creates a personalized recommendation tool right inside Claude that ends the "what should we watch?" spiral—complete with real-time searches for movies and shows that match your exact mood, plus a watchlist and group decision features.

What it looks like:

WHERE WE STAND (based on today’s stories)

✅ AI Can Now: Design new drug molecules for diseases that were previously considered untreatable.
❌ Still Can't: Make reliable medical decisions without humans double-checking its work.

✅ AI Can Now: Learn to use complicated design software by watching videos of how humans do it.
❌ Still Can't: Reliably block harmful requests when they're phrased creatively—like as poems.

✅ AI Can Now: Clone voices well enough to fool people into thinking they're talking to real government officials.
❌ Still Can't: Create performances with the emotional authenticity that comes from real human actors.

FROM THE WEB

NanoBanana is quickly becoming the do-it-all photo creation and editing tool of choice. For those of you who use Photoshop, you'll be happy to learn that Adobe is incorporating NanoBanana directly into the app instead of keeping it in the separate Firefly system.

RECOMMENDED LISTENING/READING/WATCHING

TV EPISODE: Westworld - "The Bicameral Mind" (S1E10) SPOILER ALERT!

The season one finale brings together every thread the show has been weaving. Dolores finally understands what she is, the maze reveals its true meaning, and the park's history unfolds in ways that reframe everything you watched. Anthony Hopkins delivers a farewell speech that's both chilling and oddly moving.

This episode asks what consciousness actually is and whether it matters if it emerges from code instead of biology. The reveal of Arnold's plan and Ford's counter-plan turns the whole season into a story about how AI might bootstrap itself into something more. Even if you don't continue with later seasons, this finale works as a complete statement on its own.

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.