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

A 27-Million-Parameter Model May Outperform GPT-4 and Claude

TLDR: Two 22-year-old founders turned down a multimillion-dollar offer from Elon Musk's xAI to build a brain-inspired AI architecture they believe will make the transformer era obsolete—and their 27-million-parameter prototype is already outperforming models thousands of times its size on key reasoning benchmarks.

The Story:

William Chen and Guan Wang met as high schoolers in Michigan, bonded over their shared obsession with artificial general intelligence, and ended up at Tsinghua University's brain cognition lab in Beijing—where they built one of the first language models trained with reinforcement learning. When that early project, OpenChat, caught Musk's attention and xAI came calling with a major recruitment offer, they walked away. "We decided that large-language models have their limitations," Chen told Fortune. "We want a new architecture that will overcome the structural limitation of large-scale machine learning." Their bet: a Hierarchical Reasoning Model (HRM) that mimics how the brain mixes fast reflexive thinking with slow deliberate planning. With just 27 million parameters and 1,000 training examples—no pretraining, no chain-of-thought prompting—HRM scored 40.3% on the ARC-AGI benchmark, outperforming OpenAI's o3-mini-high (34.5%), Claude 3.7 (21.2%), and DeepSeek R1 (15.8%). On tasks like Sudoku-Extreme and 30x30 maze pathfinding, where state-of-the-art models fail entirely, HRM achieved near-perfect accuracy. Their company, Sapient Intelligence, is now opening a U.S. office and preparing to scale the architecture into a general-purpose reasoning engine.

Its Significance:

If HRM's approach holds up at scale, it challenges the central assumption driving hundreds of billions in AI investment: that bigger models trained on more data will keep getting smarter. OpenAI, Anthropic, and Google are all betting on transformers. Sapient is betting they're optimizing the wrong architecture entirely—that chain-of-thought prompting is a workaround for a structural flaw, not a path to AGI. Remember what happened when DeepSeek announced they'd trained a competitive model for $5.6 million instead of hundreds of millions? Nvidia lost $589 billion in a single day—the largest market-cap wipeout in stock market history—and the Nasdaq shed $1 trillion. That was about training efficiency. If Sapient proves you need an entirely different architecture, the DeepSeek panic will look like a dress rehearsal for the real correction.

QUICK TAKES

Tesla Loses Key AI Talent to Stealth Home Robot Startup

The story: At least 10 former Tesla employees have left to join Sunday Robotics, a new startup building a wheeled home robot called Memo. The departures include engineers who spent five to seven years working on Tesla's Optimus humanoid robot and Autopilot self-driving systems. Sunday emerged from stealth with $35 million in funding and a different approach to training robots: instead of expensive VR suits in labs, it gave $200 gloves to hundreds of ordinary people who recorded themselves doing chores at home, collecting 10 million training episodes from real-world kitchens with messy counters and pets.

Your takeaway: Tesla isn't losing people to tech giants offering bigger paychecks—it's losing them to a 50-person startup with a scrappier method. The exodus suggests some of Elon Musk's top robotics engineers see more promise in simple home helpers than billion-dollar humanoids.

AI Helped Push Black Friday Online Sales to Record $11.8 Billion

The story: American shoppers spent a record $11.8 billion online on Black Friday, up 9.1% from last year. A major driver: AI shopping tools. Traffic to retail sites from AI chatbots jumped 805% compared to 2024, when tools like Walmart's "Sparky" and Amazon's "Rufus" didn't exist. Nearly half of US shoppers say they've used or plan to use AI to help with holiday shopping. Meanwhile, actual items purchased dropped slightly as higher prices from tariffs and inflation pushed costs up.

Your takeaway: AI shopping assistants went from novelty to mainstream in just one year. Shoppers are using them to find deals faster, but the real winners are retailers who can guide AI-influenced purchases their way.

Tech Giants Amass $100M+ War Chests to Fight AI Regulation

The story: Tech billionaires and AI companies are building massive political war chests ahead of the 2026 midterm elections, with spending exceeding $100 million. Super PACs backed by OpenAI's Sam Altman and Meta's Mark Zuckerberg are targeting state AI laws and backing candidates who support "lighter-touch" regulation. The industry argues a patchwork of state rules will slow innovation and hurt America's AI lead over China. Opponents, including some Republican attorneys general, say states should retain the power to protect their citizens.

Your takeaway: The fight over AI rules is moving from boardrooms to ballot boxes. With billions at stake, the outcome of the next election cycle could determine whether AI companies face meaningful oversight or mostly regulate themselves.

Microsoft Charts AI Future Less Dependent on OpenAI

The story: Microsoft CEO Satya Nadella has laid out the company's most detailed AI roadmap yet, signaling a future where Microsoft isn't tied to any single model provider. Nadella warned that AI model companies face a "winner's curse" because breakthrough models can be copied quickly. Instead, Microsoft is betting on infrastructure—identity, security, storage, and databases—as the real value in AI. The company is building massive data centers with hundreds of thousands of next-generation chips and planning for a future where AI agents, not human users, drive cloud computing demand.

Your takeaway: Microsoft helped make OpenAI what it is today, but now it's building escape routes. The message to investors: the company that controls AI's plumbing will matter more than whoever builds the best chatbot.

97% of Listeners Can't Tell AI Music from Human-Made Tracks

The story: A survey of 9,000 people across eight countries found that nearly everyone failed to tell AI-generated music from songs made by humans. Deezer, which commissioned the research, also revealed that about 50,000 fully AI-generated tracks now upload to its platform daily—34% of all new music. Meanwhile, an AI-generated country song called "Walk My Walk" by Breaking Rust topped the Billboard digital country chart, marking a first for synthetic music. Most respondents said they want AI music clearly labeled, and 70% worry it threatens musicians' livelihoods.

Your takeaway: AI music just passed a major test: it sounds real enough to fool almost everyone. The question isn't whether machines can make convincing songs—it's whether anyone will tell listeners when they do.

TOOLS ON OUR RADAR

TRENDING

Director Bong Joon-ho Wants to "Destroy AI" – The Parasite director told a Marrakech film festival jury he wants to "organize a military squad" to fight AI. Jury member Jenna Ortega added she hopes AI content becomes "mental junk food" that makes people feel sick.

Bozeman Christmas Poster Pulled Over AI Art Accusations – A $200 art contest winner in Montana had his design removed after locals accused him of using AI. The artist denies it, but the town replaced his poster with a photograph, citing safety concerns after online harassment.

AI Reshaping International Arbitration – Legal experts say AI is now transforming how disputes are resolved, from drafting documents to analyzing evidence. One organization has even introduced an AI-powered arbitrator for lower-value cases, though a human must still sign off.

Healthcare Faces AI Growing Pains – At an Arizona health summit, experts said AI could free doctors from routine paperwork—but integrating it takes one to two years and requires major changes to hospital culture and workflows.

AI Biosensors Could Speed Drug Resistance Detection – Researchers say combining AI with portable biosensors could help doctors identify antibiotic-resistant bacteria much faster than current lab methods, potentially saving lives in hospitals and remote areas.

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

Career Leverage Calculator: Map your skills, search real LinkedIn jobs, and build your career strategy—all in one interactive tool

What this does: Creates a personal career strategy tool right inside Claude that helps you see your professional value clearly, search real LinkedIn job postings matched to your skills, and build a focused action plan—like having a career coach with a recruiting assistant in your browser.

What it looks like:

WHERE WE STAND(Based on today’s Quick Takes and Trending news)

✅ AI Can Now: Generate music that 97% of listeners can't tell apart from songs made by human artists.

❌ Still Can't: Create work that carries the emotional weight of human experience—listeners who discover they've been fooled often feel tricked rather than moved.

✅ AI Can Now: Guide shoppers through complex purchases using natural conversation, remembering preferences and comparing prices across thousands of products.

❌ Still Can't: Replace the instinct of physically handling a product or asking a knowledgeable store employee follow-up questions in real time.

✅ AI Can Now: Learn household tasks from videos of regular people doing chores in messy, unpredictable home environments with pets, clutter, and poor lighting.

❌ Still Can't: Reliably walk on two legs—which is why some robotics teams are skipping humanoid designs entirely and putting their robots on wheels instead.

FROM THE WEB

RECOMMENDED LISTENING/READING/WATCHING

PODCAST EPISODE: Hard Fork - "The AI Arms Race"

Kevin Roose and Casey Newton break down the competition between OpenAI, Google, and Anthropic without drowning you in tech jargon. They explain why these companies are burning billions of dollars and racing to build smarter AI, and what it means when the people building this technology can't agree on whether it's safe.

The episode balances humor with genuine concern. Roose and Newton have a chemistry that makes complex topics digestible, and they're not afraid to ask the uncomfortable questions about where this race leads. Perfect for understanding why AI suddenly feels like it's everywhere and why that matters.

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.