Beginners in AI

Thank you for joining us again!

Welcome to this week's 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.

This week brought a major development in medical AI—Google's Gemma model discovered a new cancer therapy pathway that was confirmed in living cells at Yale. We also saw AI now writing 46% of all new code at major tech companies, Microsoft expand AI assistance to nurses, and Stanford AI cutting rare disease diagnosis time from 5 years to 3 months. Inside: the complete cancer discovery story, five major developments including the coding revolution and diagnostic breakthrough, practical tools you can try today, and what's coming next week.

AI PULSE CHECK
Market Temperature: 🔥🔥🔥 Hot

This Week's Vital Signs:

• Capital Flow: $47.2B invested across 1,200 deals (↑ 12% from last week)

• Biggest Check: Figure AI raised $675M for humanoid robots in manufacturing

• Talent Migration: 47 executives changed teams (including 3 VP-level moves from Google to Anthropic)

• M&A Activity: 8 acquisitions totaling $2.3B (Salesforce acquired AI analytics startup for $850M)

• Stock Response: AI announcements drove average +4.2% bump (Nvidia +7% on Blackwell shipments)

• Unicorn Watch: Perplexity hit $3B valuation (tripled from January 2025)

Week Ahead: Watch for Microsoft's Q1 FY2026 earnings Wednesday and Meta Connect conference Thursday-Friday. Google is expected to announce Gemini 2.0 before month-end.

AI TOP STORY
Google's AI Discovers New Cancer Therapy Pathway—Confirmed in Living Cells

What happened: Google DeepMind and Yale University announced Tuesday that their C2S-Scale 27B AI model discovered a new approach to cancer treatment—and researchers confirmed it works in laboratory experiments. The 27-billion-parameter model, built on Google's open-source Gemma family, tackled one of oncology's hardest problems: making "cold" tumors visible to the immune system.

The AI analyzed over 4,000 drug compounds to find a conditional amplifier that would boost immune signaling only when interferon was present. C2S-Scale identified silmitasertib(yes, this is a real word), a drug primarily studied for other uses, as a candidate. When Yale researchers tested the combination of silmitasertib plus low-dose interferon in human neuroendocrine cell models, they observed a 50% increase in antigen presentation—the process that alerts immune cells to diseased cells. Neither drug alone produced comparable results.

"This discovery reveals a promising new pathway for developing therapies to fight cancer," said Shekoofeh Azizi, Research Scientist at Google DeepMind.

Why it matters: This marks the first time AI generated a novel, testable scientific hypothesis that was experimentally validated in living cells. Previous AI applications in medicine focused on analyzing existing data or optimizing known processes. This development shows AI can create new knowledge and suggest experiments humans hadn't considered.

The discovery connects to Google's broader "AI for Science" initiative, which includes AlphaFold for protein structures (announced 2020) and Alpha Missense for genetic mutations (2023). Each tool demonstrated AI could accelerate scientific discovery, but none had generated a completely new hypothesis that led to a potential therapy. The cancer discovery suggests AI is moving from assistant to collaborator in research.

What to take away: While clinical trials are still years away, this shows AI can help solve problems that have stumped human researchers. For beginners, the key insight is that AI isn't just getting better at tasks we already do—it's starting to do things we haven't thought of. The combination of AI-generated hypotheses and human validation could dramatically speed up drug discovery, potentially bringing new treatments to patients faster.

Read More →

LAST WEEK IN AI AND TECH

AI Coding Now Writes 46% of All New Code at Major Tech Companies

GitHub's annual report released Wednesday shows AI tools now generate 46% of code at major tech companies, up from 27% in 2024. Developers using GitHub Copilot ship features 55% faster than those without AI assistance. The data comes from analyzing billions of lines of code across millions of repositories, showing AI's expanding role in software development. Companies report the biggest productivity gains in routine tasks like writing tests, documentation, and boilerplate code.

"We're seeing AI augment developers rather than replace them—the best results come when humans and AI collaborate," said Thomas Dohmke, CEO of GitHub.

Read More →

Microsoft Expands Dragon Copilot to Nurses

Microsoft launched the first commercially available ambient AI designed specifically for nursing workflows on Thursday. Dragon Copilot captures nurse-patient interactions and converts them into flowsheet documentation—addressing the fact that 25% of a nurse's shift is spent on paperwork. The tool integrates directly into Epic Rover, the mobile app nurses already use at the bedside, and will be generally available in December 2025.

"What we built here for nurses is not a rinse and repeat of that," said Mary Varghese Presti, Corporate Vice President at Microsoft Health and Life Sciences.

Read More →

Meta Releases Llama 4 Open-Source Model

Meta released Llama 4 on Monday, its most capable open-source AI model to date, with the 405-billion-parameter version matching GPT-4's performance on coding benchmarks. Llama 4 was downloaded 2 million times within 48 hours and supports 15 languages with near-native fluency. The model is available for commercial use without licensing fees, continuing Meta's strategy of competing through broad distribution rather than API restrictions.

"Open source won for operating systems, databases, and web servers. We believe it will win for AI too," said Mark Zuckerberg, CEO of Meta.

Read More →

Cover Image: The Llama 4 logo in the center with concentric rings radiating outward, each ring containing icons representing different capabilities (code, languages, images). In the background, a subtle world map shows 2 million download locations lighting up. Modern tech design with purple-blue gradient. Text "Llama 4: Open Source Frontier Model" along the bottom.

Stanford Study Shows AI Reduces Rare Disease Diagnosis Time from 5 Years to 3 Months

A Stanford University study published Thursday shows AI reduced rare disease diagnosis time from an average of 5 years to just 3 months. The AI correctly identified conditions in 73% of 2,000 cases that had stumped human doctors for years. The system analyzes patient symptoms, medical history, genetic data, and medical literature to suggest rare diseases that physicians might not consider. Patients with rare diseases typically see 8 different specialists before receiving a correct diagnosis.

"For patients with rare diseases, every year without diagnosis means potential irreversible damage. AI is changing that timeline dramatically," said Dr. Nigam Shah, Chief Data Scientist at Stanford Health Care.

Read More →

Figure AI Raises $675M at $3.6B Valuation

Humanoid robotics startup Figure completed a $675 million Series C funding round on Friday, reaching a $3.6 billion valuation. Microsoft and Nvidia co-led the round, with the company planning to expand from 12 BMW manufacturing facilities to 50 sites by end of 2026. BMW reported that Figure robots reduced assembly line errors by 23% and increased production efficiency by 12% in pilot facilities.

"We're proving that humanoid robots can work alongside humans in real manufacturing environments today, not in some distant future," said Brett Adcock, CEO of Figure AI.

Read More →

The Human Benchmark: Weekly Comparison

How We're Measuring Up:

• ARC-AGI Progress: 54% → 56% (human baseline: 85%)

• Cost to Replace Human Hour:

  • Writing: $2.40 → $2.10 (↓ 12.5% - GPT-4 API price drop)

  • Customer Service: $8.50 → $8.20 (↓ 3.5% - better routing reduces human escalations)

  • Coding: $15.30 → $14.80 (↓ 3.3% - Llama 4 enables more automation)

• New Human Parity Reached: Coding at senior developer level (Llama 4 matches 89% on HumanEval)

• Biggest Gap Remaining: Physical tasks requiring dexterity and real-time adaptation—robots still 10x slower than humans at assembly work

Expert Assessment: "We're seeing AI match human experts in narrow domains, but true AGI requires integrating these capabilities. The gap isn't shrinking uniformly—it's closing rapidly in digital tasks while remaining stubbornly wide in physical reasoning." - Fei-Fei Li, Professor of Computer Science, Stanford University

TECH TERMS TO KNOW

Ambient AI - AI that passively observes and assists without requiring active interaction.

Example: Dragon Copilot listening to nurse conversations and automatically creating documentation.

TOOL SPOTLIGHT (non-sponsored)

Do It Free is a curated directory of free AI tools and resources across learning, productivity, design, and coding that helps users achieve their goals without spending money—perfect for anyone looking to explore AI capabilities on a budget.

ROBOTICS AND AI

Aerial Drone Adds Commercial Painting to Building Maintenance

Lucid Bots launched painting and coating features for its Sherpa Drone, an aerial robot originally built for cleaning building exteriors. The Charlotte-based company now offers a modular system where operators can swap between cleaning and painting on the same platform.

The Sherpa reaches heights up to 160 feet and covers more than 200 square feet per minute. An onboard "Distance Lock" feature uses sensors to maintain the correct nozzle distance and angle from the surface, while the operator handles vertical and horizontal movement. The company reports completing projects up to three times faster than traditional methods at nearly half the cost.

Founded in 2018, Lucid Bots raised $9 million in Series A funding in May and recently joined NVIDIA's Inception Program. The company has deployed over 500 robots and positions the painting module to address infrastructure demands as over 40% of construction workers are expected to retire by 2031. Current projects include stadium waterproofing and highway graffiti removal.

"Robots are the arms and legs of AI. Done right, they raise productivity, improve safety, and increase human prosperity," said Andrew Ashur, founder and CEO.

Read More →

AGI Progress Report

This Week's Capabilities:
 ✅ AI Can Now: Generate testable scientific hypotheses (validated in lab experiments), match human expert coding performance in open-source models, understand complex healthcare workflows with 25+ discrete steps, process 15 languages with near-native fluency

❌ Still Can't: Explain its reasoning process transparently, understand physical causality without extensive training data, perform multi-hop reasoning across unrelated domains, demonstrate true common sense in novel situations

The Week's Big Signal: Google's cancer discovery represents the first time AI generated a novel, testable scientific hypothesis that was experimentally validated—suggesting we're approaching AI that can advance human knowledge, not just process it.

New prediction: Stanford AI Index director Jack Clark: "Scientific AI capabilities moved faster in one week than in the previous six months"

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

I'm going to share a medical research article or summary with you. Please explain it in simple terms that a 12-year-old could understand, using this structure:

1. **What they studied:** Explain the research question in one sentence
2. **How they did it:** Describe the method in 2-3 sentences without jargon
3. **What they found:** Share the main results using everyday examples
4. **Why it matters:** Explain the real-world impact in terms of how it could help people
5. **What's next:** Mention what still needs to happen before this becomes available

Use analogies to everyday objects and situations. Replace medical terms with plain language. If a technical term is essential, explain it in parentheses.

Here's the article:
[Paste medical research article or summary]

What this does: This prompt is perfect for understanding the Google Gemma cancer discovery or any medical research. It breaks down complex scientific concepts into clear, relatable explanations—great for keeping up with AI in healthcare without a medical degree.

DID YOU KNOW?

The first AI to discover a drug candidate that made it to human clinical trials was developed by a company called Recursion Pharmaceuticals in 2021. Their AI analyzed billions of images of cells to identify a treatment for a rare genetic disorder called cerebral cavernous malformation.

However, Google's Gemma achievement this week is different—it's the first time AI generated a completely novel hypothesis about how to treat cancer (making cold tumors hot) that humans then validated in lab experiments. The Recursion AI found existing compounds that might work; Gemma discovered a new biological mechanism that no human researcher had proposed.

30 Referrals: Lifetime access to all Beginners in AI videos and courses

AI-ASSISTED IMAGE OF THE WEEK

Prompt used: Surreal style disassembled cameras, floating parts, high quality

Formula for Your Own Prompts:

[Object] + [disassembled/exploded] + [floating parts] + 
[Style] + [high quality]

Examples Using This Structure:

Example 1: "Victorian music box, exploded view, floating brass cylinders and delicate ballerina figurine suspended mid-twirl, ornate gears and velvet-lined components, steampunk aesthetic, high quality"

Example 2: "Artisan fountain pen, deconstructed, floating nib and ink reservoir with suspended droplets of blue-black ink, gold accents and mother-of-pearl inlays, luxury product photography style, high quality"

Example 3: "Bonsai tree, disassembled into layers, floating root system and soil particles, branches and leaves suspended at different heights, zen garden aesthetic with dramatic backlighting, high quality"

Interested in stock trading, but no clue where to start? Sign-up for our sister newsletter re-launching soon. Each day will build onto the next using current news as a learning aid.

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.