Sponsored by
Your Kafka Bill Is an Architecture Problem
More than 80% of Kafka costs aren't hardware – they're interzone networking fees. WarpStream BYOC eliminates them entirely by replacing stateful brokers with stateless agents that write directly to your own object storage.
No disks, no inter-AZ replication, no partition fees. Goldsky cut TCO by 90%+. Your existing clients keep working – just point them at a new URL.
Learn how it works, then sign up for free. Get $400 in credits that never expire. No credit card required to start.
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 edited, and published with the intention of making AI news and technology more accessible to everyone.
THE FRONT PAGE
AI Deepfakes Were a Threat. Now YouTube Wants You to Make Your Own
TLDR: YouTube just launched a tool that lets any channel owner create short videos starring a realistic AI copy of their own face and voice.
The Story:
Google rolled out a new YouTube Shorts feature called "Make a video with my avatar," powered by its Veo 3.1 video model. Any YouTube channel owner aged 18 or older can now generate short clips starring a digital version of themselves, just by typing a text prompt. Each clip runs up to about eight seconds, but users can chain clips together for longer videos. The feature is currently only available in YouTube Shorts and the YouTube Create app. A YouTube spokesperson told Decrypt the tool "gives users an easier way to include themselves safely and securely in videos." CEO Neal Mohan had flagged this as a 2026 priority earlier this year, specifically calling out avatar-based Shorts as part of a broader push to expand AI creation tools for creators.
Its Significance:
This is a big deal for anyone who makes videos but doesn't always want to be on camera. You could film a video of yourself without actually filming yourself. But it also raises real questions about trust: if anyone can make a realistic video of themselves saying anything, how will viewers know what's real? The same week YouTube launched this creator tool, it was also expanding a separate system to detect and remove fake AI videos of politicians and journalists. These two moves are happening at the same time, and they point in opposite directions. More AI creation tools are coming to the masses, and the space between "me" and "an AI playing me" is about to get weird.
QUICK TAKES
The story: OpenAI is developing a restricted cybersecurity AI product called "Trusted Access for Cyber," just days after Anthropic revealed its own locked-down model, Claude Mythos Preview. Mythos was already alarming: in testing it found thousands of previously unknown software bugs across every major operating system and browser, scored a perfect 100% on a top cybersecurity benchmark, and successfully built working exploits for Firefox 84% of the time.
Your takeaway: Both OpenAI and Anthropic now have AI systems so good at hacking that they won't release them to the general public. Instead, they're offering access only to vetted companies like Microsoft, Apple, and CrowdStrike to help find and fix bugs before bad actors do. That's a first for the AI industry, and it signals that the most powerful AI tools are quietly being locked away from everyday users.
The story: Tether, the company behind the world's biggest digital dollar (USDT), launched a free, open-source toolkit called QVAC SDK that lets developers build AI apps that run entirely on your device, with no internet connection needed. It works across iPhone, Android, Windows, Mac, and Linux, and supports text generation, voice, translation, and image recognition.
Your takeaway: This is a privacy play wrapped in a developer tool. AI that runs locally means your data never leaves your phone. For people in areas with spotty internet, or anyone wary of sending their data to big tech servers, this is the kind of AI they've been waiting for. Tether is betting that the future of AI is offline, personal, and decentralized.
The story: A team of researchers from Microsoft, Google DeepMind, Columbia University, and two AI startups published a proposal called the Agentic Risk Standard (ARS), a framework for protecting people when an AI agent makes a bad financial move. The idea borrows from how traditional finance handles risk: escrow, insurance, and collateral. In simulations across 5,000 trials, the system reduced user losses by up to 61%.
Your takeaway: AI agents are already making trades, sending payments, and handling money on behalf of real people. But when one of them messes up, there's currently no safety net. This research is one of the first serious attempts to build one, and it matters because the stakes are real. Expect this kind of framework to become a hot topic as more AI agents get handed the keys to people's finances.
TOOLS ON OUR RADAR
Jellyfin Free and Open Source: A personal media server software that lets you organize your own movie and music collections and stream them directly to your television or mobile devices seamlessly. (Alternative to Plex)
🌀 IllusionDiffusion Free: A creative web application that uses advanced control networks to generate stunning high quality optical illusions by blending your custom text prompts with hidden patterns or images.
📈 Scenario Edge Freemium: An intelligent financial platform that helps investors and portfolio managers analyze their holdings using scenario based simulations to stress test against custom macroeconomic events and market shocks.
👤 Personify Freemium: A sophisticated coaching platform that allows course creators to build a personalized artificial intelligence clone trained on their own videos and frameworks to deliver automated student support around the clock.
TRENDING
OpenAI Has a New $100 ChatGPT Pro Plan to Better Match Up With Claude — OpenAI launched a new $100/month ChatGPT Pro tier that sits between its existing $20 Plus and $200 Pro plans. It's specifically aimed at heavy Codex users, offering 5x more coding capacity than the Plus plan. The move is a direct shot at Anthropic's Claude, which already offered a matching $100/month tier. OpenAI says Codex usage has grown 70% month-over-month and now has 3 million weekly users.
Google's PaperOrchestra AI Converts Lab Notes Into Publication-Ready Research Papers — Google researchers unveiled PaperOrchestra, a five-agent AI system that turns rough lab notes and experimental data into finished academic papers, complete with literature reviews, figures, and citations. In human evaluations it outperformed other AI writing tools by up to 68% on literature quality. The full process takes under 40 minutes.
Google News Now Prominently Featuring Polymarket Bets Instead of Actual Journalism — Google News has started surfacing Polymarket gambling odds alongside real news stories, and in some cases the betting market results appear at the very top of the page. Users can even set Polymarket as a news "source" in Google's search bar. Critics say this blurs the line between journalism and speculation, while Polymarket has faced scrutiny for allowing insider trading on major news events.
Google Launches Notebooks in Gemini, Synced With NotebookLM — Google rolled out a new "notebooks" feature inside the Gemini app that acts like a personal knowledge base. You can drop files, PDFs, and past chats into a notebook, give Gemini specific context, and it automatically syncs with NotebookLM, giving you access to that app's video overviews and infographic tools. It's available now for paid Google AI subscribers, with free access coming soon.
Tesla Robotaxi App Gets Update With Interior Camera Safety and Pulsing Lights — Elon Musk pushed fans to download the updated Tesla Robotaxi app, which includes new safety tips, a refreshed interface, and clearer instructions for riders. Key updates: vehicles seat up to four passengers, exterior lights pulse on arrival, there's a 7-minute pickup window, and interior cameras may monitor the cabin for safety. Tesla's robotaxi service is currently running in Austin and parts of the San Francisco Bay Area.
OpenAI Backs Bill That Would Exempt AI Firms From Lawsuits Over Mass Harm — OpenAI testified in support of an Illinois bill, SB 3444, that would shield AI companies from lawsuits in cases of "critical harm" like mass casualties or $1 billion in property damage, as long as the company wasn't intentionally reckless and had published safety reports. The bill applies to any AI model trained on more than $100 million in compute, which would cover OpenAI, Google, Anthropic, xAI, and Meta. Critics say it's a more extreme liability shield than anything the industry has previously pushed for.
TRY THIS PROMPT (copy and paste into Claude, ChatGPT, or Gemini)
🛠️ Build a meeting cost calculator that shows the real dollar value of every meeting based on attendees, hourly rates, and duration
Build me a single-file HTML meeting cost calculator using React 18 (CDN) and Babel standalone. Features: 1. Live timer that counts up in mm:ss and shows real-time cost (people x rate x elapsed hours). Start/Pause/Reset buttons. 2. Meeting stats panel showing total spent, average cost, meeting count, and most expensive meeting. 3. Log a meeting form with name, attendees (number), average hourly rate, duration in minutes, and meeting type dropdown (Standup, Planning, Review, Interview, Workshop, All-hands, One-on-one, Other). Show projected cost before logging. 4. Meeting log on the right with color-coded entries by type, showing cost bars relative toWhat this does: A live meeting cost tracker with a running timer. Enter your headcount and average hourly rate, hit Start, and watch the dollar amount tick up in real time. Log past meetings to see your total spend and spot which meeting types cost the most.
What this looks like:
WHERE WE STAND(based on today’s news)
✅ AI Can Now: Create a realistic short video starring your face and voice from a simple text prompt, without you ever appearing on camera.
❌ Still Can't: Tell the difference between a real video and an AI-generated avatar, without special detection tools that most people don't have access to.
✅ AI Can Now: Find decades-old software bugs that human security teams missed, at a scale and speed no human expert could match.
❌ Still Can't: Be trusted to execute financial trades without a safety net. Researchers found losses still occur 39%–76% of the time without proper underwriting in place.
FROM THE WEB
RECOMMENDED LISTENING/READING/WATCHING
How Machines Learn by CGP Grey (YouTube Video)
CGP Grey explains machine learning through the lens of evolution, showing how AI systems improve through trial, error, and selection, just like biological organisms. His signature stick-figure animations make complex concepts click instantly. If you only watch one explainer about how AI actually works, make it this one.
Hiring in 8 countries shouldn't require 8 different processes
This guide from Deel breaks down how to build one global hiring system. You’ll learn about assessment frameworks that scale, how to do headcount planning across regions, and even intake processes that work everywhere. As HR pros know, hiring in one country is hard enough. So let this free global hiring guide give you the tools you need to avoid global hiring headaches.
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.
Some * designated product links may be affiliate or referral links. As an Amazon Associate, I earn from qualifying purchases. This helps support the newsletter at no extra cost to you and Amazon makes a tiny hair less.
