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

DeepMind Cracked the Navier-Stokes Problem

TLDR: Google DeepMind used AI to discover new mathematical solutions to centuries-old fluid equations, working with researchers from Brown, NYU, and Stanford.

The Story:

Google DeepMind just did something different. They didn't use AI to calculate faster. They used it to find answers that humans never found in the first place.

The team built a special kind of AI called physics-informed neural networks—basically, AI that's trained to respect the laws of physics while it works. Instead of learning from massive datasets, this AI learns by checking itself against actual physics equations. When applied to fluid dynamics, it discovered solutions that were hiding in plain sight.

Getting there took extreme precision. The accuracy required was so fine it's like predicting Earth's diameter to within a few centimeters. Why? Because they were hunting for singularities—theoretical points where fluid speeds or pressures jump to infinity. These are rare, fragile mathematical objects that barely exist. Finding them matters because they reveal what breaks in our equations.

Why This Matters:

Solving the Navier-Stokes equations—the equations that describe how liquids and gases move—is one of only six unsolved Millennium Prize Problems. Win? You get $1 million. Mathematicians have been stuck on this for over 200 years.

What makes it so hard: these singularities, if they exist, only show up under conditions that are almost impossibly precise. Most AI and math tools find "stable" solutions that are robust. But the important ones are unstable—fragile, elusive, and nearly impossible to spot with traditional methods.

DeepMind's approach was smart. Instead of just throwing computing power at the problem, they wove mathematical insights directly into how the AI learned. They used better optimization techniques and pushed the AI to near-perfect precision. The result? New families of singularities that mathematicians can now study and verify.

The Real Shift:

This is about more than solving one equation. The way they worked together—AI discovering, humans proving—might be how science works going forward. Instead of AI just speeding up human work, it becomes a research partner that finds things we'd never think to look for.

Practical payoff? Better aircraft design, weather forecasts, and models for how water moves through soil. Better understanding of ocean currents and turbulence. But the bigger story is the method itself: we're learning to use AI for actual discovery, not just computation.

QUICK TAKES

Tiny Brain Implants Could Travel Through Your Bloodstream Without Surgery

The story: MIT researchers developed microscopic wireless implants smaller than a grain of rice that can travel through the bloodstream and self-implant themselves in the brain. The chips are integrated with immune cells so the body doesn't reject them, and they can deliver precise electrical treatment to target brain inflammation. The technology could eventually treat deadly diseases like brain tumors, Alzheimer's, and multiple sclerosis without requiring dangerous brain surgery.

Your takeaway: Doctors may soon be able to treat serious brain diseases with a simple injection instead of invasive surgery, making life-saving treatments available to far more people.

Google Releases Faster, Smarter File Detection with Magika 1.0

The story: Google launched Magika 1.0, an AI-powered tool that identifies file types with improved performance and security. The system has already been downloaded over 1 million times since going open-source last year, and Google rewrote the core engine in a faster programming language called Rust.

Your takeaway: Developers now have a faster, more reliable way to sort through files using AI, which could make software tools work better and more securely.

Google's Gemini AI Gets "Deep Research" to Find Your Stuff

The story: Google added a new "Deep Research" feature to Gemini that lets it search through your personal data. It can look at your emails, files, slides, and chat history to give you more personalized and accurate results, all while keeping your information private.

Your takeaway: AI can now understand your personal context better, making it a more useful research assistant that feels like it actually knows your specific needs.

OpenAI's Sora Video Tool Comes to Android

The story: OpenAI's Sora app, which creates AI-generated videos from text descriptions, is now available on Android phones in the US, Canada, Japan, South Korea, Taiwan, Thailand, and Vietnam. Until now, the app only worked on iPhones, so this opens the door to billions more users since Android phones make up about 70% of the global smartphone market.

Your takeaway: AI video creation is becoming more accessible to mainstream audiences globally, not just those with expensive phones.

Google Confirms Gemini 3 Launch Coming Later This Year

The story: During an earnings call, Google's CEO Sundar Pichai announced that Gemini 3, the next major AI model, will launch later in 2025. The new version will focus on "agent-like" capabilities, meaning it can handle complex tasks that involve multiple steps and different types of information at once.

Your takeaway: Google is pushing hard to catch up with competitors like OpenAI by building AI that can tackle more complex, real-world problems on its own.

Microsoft Launches Its Own AI Image Generator: MAI-Image-1

The story: Microsoft introduced MAI-Image-1, its first homemade AI image generator, now available in Bing Image Creator and Copilot. The tool excels at creating images of food, nature scenes, and lighting effects, and Microsoft designed it to be fast and high-quality so creative workers can use it throughout their day.

Your takeaway: Big tech companies are building their own AI image tools instead of relying on others, giving them more control over what their users can create.

1Mind Raises $30 Million to Deploy AI Sales Agent "Mindy"

The story: 1Mind, founded by a former 6Sense executive, just raised $30 million to expand its AI sales agent named Mindy. Unlike other AI sales tools that send emails and make cold calls, Mindy handles customers who are already interested, helping them through the entire buying process and even standing in for sales engineers on important calls. The company is already working with major firms like HubSpot and LinkedIn.

Your takeaway: AI is moving beyond simple outreach tasks to handle the complex, high-value work of closing deals and managing customer relationships.

AgiBot Achieves Industrial Breakthrough: Robots Learn New Factory Tasks in Minutes

The story: Chinese robotics startup AgiBot successfully deployed AI-powered robots on a real manufacturing line that can learn new tasks in minutes without needing new code or reprogramming. When the production model changed, the robots simply retrained in minutes and kept working. This is the first time reinforcement learning—a type of AI that learns by doing—has worked in actual factory settings, not just labs.

Your takeaway: Factories could soon have robots that adapt on the fly instead of needing engineers to reprogram them for every small change, making production faster and cheaper.

WHERE WE STAND

✅ AI Can Now: Generate video and images directly from simple text descriptions that rival human creativity.

❌ Still Can't: Handle complex tasks without being retrained when conditions change—robots on factory floors still struggle with flexibility. That may change with AgiBot's new announcement.

✅ AI Can Now: Learn new manufacturing tasks and adapt to production changes in minutes instead of requiring days of programming.

❌ Still Can't: Reliably handle high-stakes decisions like closing major sales deals without a human involved—companies still want people involved in their biggest decisions.

✅ AI Can Now: Reach inside the body to deliver precise medical treatment and stimulation without the need for surgery.

❌ Still Can't: Prevent the body's immune system from rejecting foreign materials, though new biocompatible designs are getting closer.

TOOLS ON OUR RADAR

• 🔨 Factiverse [Freemium]: AI fact-checker that instantly verifies claims in text, video, and audio across 100+ languages, helping journalists and organizations fight misinformation in real time.

• 📐 Hypotenuse AI [Freemium]: Generate SEO-optimized product descriptions, blog articles, and marketing copy in seconds using AI trained on high-quality data for authentic-sounding content.

• 🔧 Udio [Freemium]: Create full songs with AI vocals, instrumentals, and lyrics from simple text descriptions—no musical skills required, with custom controls for professionals.

• 🛠️ Atlassian Confluence[Paid]: Centralized team workspace for real-time document collaboration, project management, and knowledge sharing with AI-powered creation and thousands of integrations.

• 🪛 Audacity [Open Source & Free]: Free, multi-track audio editor for recording and editing podcasts, music, and voice-overs with built-in effects and VST plugin support.

• 🧰 Libation [Open Source & Free]: Download your Audible audiobook library to DRM-free MP3 or M4B files, organize by metadata, split chapters, and keep them forever on any device.

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

You're a video production assistant helping create a 60-second product demo script. I'm launching [PRODUCT NAME] and need a script that:

- Opens with a clear hook (show the problem being solved in the first 5 seconds)
- Demonstrates 3 key features with specific on-screen text labels
- Includes natural transitions between each feature
- Closes with a clear call-to-action
- Reads naturally when spoken aloud (avoid corporate jargon)
- Includes timing cues and on-screen text suggestions

The target audience is [DESCRIBE YOUR AUDIENCE]. The tone should feel [HELPFUL/PLAYFUL/PROFESSIONAL].

Output the script in a two-column format:
LEFT COLUMN: Voiceover script
RIGHT COLUMN: Visual directions (what appears on screen, text overlays, transitions)

Make each section take approximately 15-20 seconds.

What this does: Creates production-ready scripts for demo videos that work with tools like Veo 3.1, Runway, or traditional video editors. Generates clear visual direction that non-technical creators can follow.

FROM THE WEB

Google Flow is introducing camera controls as a core prompting feature for AI video generation. Rather than requesting abstract transformations, the tool encourages users to think like directors—specifying physical camera movements like whip pans and tracking shots to choreograph scene transitions.

While video generation with Veo is available directly from your Gemini interface alongside text generation at https://gemini.google.com/, to access the advanced camera control features requires using Veo on the Google Flow platform instead at https://labs.google/flow/about

RECOMMENDED READING (to aid with Google Flow’s new features)

TRENDING

💻 "Vibe Coding" Goes Mainstream

Collins Dictionary just named "vibe coding" its Word of the Year 2025—a signal that AI-assisted programming crossed from tech circles into everyday vocabulary. The term, coined by OpenAI co-founder Andrej Karpathy, means "telling a machine what you want instead of painstakingly coding it yourself." Translation: natural language → working code via AI. It jumped from zero usage in January to "huge increase" by November. Why it matters: coding just became accessible to non-engineers. The barrier to app-building collapsed. Look for voice interactions to be the new norm across the board.

🏭 Cognizant Deploys Claude to 350,000 Staff (Agents Going Real)

Cognizant announced November 4 it's rolling out Anthropic's Claude across 350,000 employees globally—not a pilot, actual production deployment. The company is embedding Claude, Claude Code, and agentic tools into its engineering platforms to move clients from "AI experimentation to scaled business outcomes." Key: using Claude for software modernization, legacy code refactoring, testing, and multi-step workflows with human oversight. This is the pattern: agencies testing agents at 350K+ scale. If it works, expect cascade of copycat deployments.

🔒 Cybersecurity Becomes Existential (AI on Both Sides)

Chinese hackers broke into America's phone networks and listened to calls from top government leaders in what Senator Mark Warner called the "worst telecom hack in U.S. history". Salt Typhoon accessed call records and messages from over a million users, including presidential staff. The incident proves AI is now both weapon and defense: attackers use AI for sophisticated social engineering; enterprises pour money into AI-powered threat detection. Cybersecurity jumped from "important" to "existential." Budget increases expected through 2026.

🧪 Teen Safety Becomes AI Product Priority

OpenAI released its Teen Safety Blueprint November 8, a product and policy playbook putting teen well-being at center stage. Features include age-appropriate design, parental controls with notifications, and age prediction so under-18 experiences tune by default. Why it matters: AI companies are treating safety as first-class feature, not cleanup. This becomes table stakes for enterprise adoption in regulated industries. Expect compliance functions to demand this across all AI vendor contracts by 2026.

📊 OpenAI's New "Tasks" Feature (Agents Shipping)

OpenAI unveiled Tasks, a new feature for AI-driven virtual assistants announced November 8—agents that can trigger multi-step workflows without human input at each step. Combined with Cognizant's 350K deployment and Anthropic's Agent SDK gaining traction, autonomous agents are crossing from concept to production deployment this month. Markets will watch: if agents prove reliable at 350K+ scale, capital floods into agent infrastructure and DevTools.

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.