Tuesday, November 11, 2025
Happy Tuesday!
Welcome to another issue of HOW TO // AI. We’re glad you made it!
Here’s what we’ve got for ya:
🤖 How AI Chatbots Really Work
🤓 OpenAI Leader Expects AI To Oust Interns
💬 ChatGPT Can Now Record Meeting Summaries
🗣️ Quick Hits: Policy & Governance, Product Upgrades, Content & Discovery
Let’s get started!
Shoppers are adding to cart for the holidays
Over the next year, Roku predicts that 100% of the streaming audience will see ads. For growth marketers in 2026, CTV will remain an important “safe space” as AI creates widespread disruption in the search and social channels. Plus, easier access to self-serve CTV ad buying tools and targeting options will lead to a surge in locally-targeted streaming campaigns.
Read our guide to find out why growth marketers should make sure CTV is part of their 2026 media mix.
🤖 How AI Chatbots Really Work
It Eats the Internet (But Doesn’t Memorize It)
Imagine feeding a robot millions of books, websites, and articles — like giving it the world’s biggest library. It doesn’t remember every word. Instead, it learns patterns 😀 😀
“After ‘Happy’, the next word is often ‘birthday’ or ‘day’.”
Words → Puzzle Pieces (Tokens)
Every word gets chopped into tiny bits called tokens. Even punctuation counts!
Text --> Becomes Tokens
Hello, World! --> ["Hello", ",", " world", "!"]These get turned into secret numbers only the AI understands.
It Plays "Guess the Next Word" — Billions of Times
The AI looks at everything you’ve typed so far…
…then asks:
“What word comes next?”
It guesses. You type more. It guesses again. Like super-smart autocomplete on your phone, but 1000x better.
Humans Train It to Be Helpful & Safe
After the robot learns patterns, people step in:
“Don’t say mean things.”
“Explain math step-by-step.”
“Be funny, not boring.”
They use examples, rules, and your feedback to shape its personality.
Now You Can Chat, Write, or Build!
The same “guess next word” trick powers:
✅ Writing emails
✅ Answering questions
✅ Coding help
✅ Jokes, stories, translations
📌 Key Takeaway
AI doesn’t know facts like a human.
It’s a pattern-predicting word machine trained on the internet — and it builds sentences one guess at a time.
Want a Fun Analogy?
Think of an LLM like a chef who’s tasted every recipe online, but never went to school.
You say: “Make me chocolate cake.”
It doesn’t look up a recipe — it just remembers the pattern of ingredients and steps from a billion cakes… and cooks one from scratch!
The Rundown: OpenAI CEO Sam Altman predicts that AI will soon match the skills of junior employees and eventually rival experienced engineers, raising concerns about the future of entry-level jobs.
Details:
Altman expects AI agents to start solving complex business problems autonomously within a year.
Despite these warnings, Gen Z is adopting AI at twice the rate of older generations, often viewing it as a co-worker or assistant.
Tech leaders like Nvidia’s Jensen Huang warn that those who fail to embrace AI risk being outpaced in the job market.
Takeaway: AI is taking over the repetitive tasks that used to fall to entry-level hires. That shift is forcing companies to rethink how they build their teams. Instead of hiring to compete with AI, it’s time to train people to work with it, starting with junior talent.
The Rundown: OpenAI has launched new business-focused features for ChatGPT, including connectors for popular cloud services and meeting recording capabilities.
Details:
These integrations allow users to search documents and convert meeting transcripts into actionable items.
ChatGPT now connects with Dropbox, Box, SharePoint, OneDrive, and Google Drive, enabling users to analyze information across these services.
The new meeting recording and transcription feature generates time-stamped notes, suggests actions, and allows users to convert action items directly into applications.
Takeaway: ChatGPT’s new enterprise tools make it easier to organize business data and pull out useful insights. They help simplify research, planning, and follow-up for teams of any size.
Policy & Governance 🏛️
⚖️ OpenAI pressed back against data preservation order in the NYT copyright case, arguing it conflicts with user privacy commitments.
🎓 Chinese AI tools went dark during nationwide college exams as AI programs disabled image recognition nationwide to keep students from cheating.
🔒 Anthropic unveiled "Claude Gov" models for US national security with enhanced capabilities for handling classified materials.
Product Upgrades 🚀
🧠 Google rolled out a major update to its Gemini 2.5 Pro model aimed at fixing previous issues and delivering better formatting of creative responses.
🗣️ ChatGPT's voice mode received an upgrade with more natural-sounding speech featuring subtler intonation, realistic cadence, and better expressiveness.
Content & Discovery 📚
📰 The New York Times inked a multiyear AI licensing deal with Amazon to bring its journalism to various AI-powered products.
📜 AI analysis suggested the Dead Sea Scrolls may be older than previously thought. The program "Enoch" was specially developed to analyze the scrolls.
🔊 HOW AI Works - Chatbots and LLMs
This video is an excellent starting point for AI beginners because it clearly explains how chatbots and large language models like ChatGPT actually work—breaking complex ideas into simple, visual concepts. It connects everyday chatbot interactions to the underlying neural networks, training data, and response generation processes without overwhelming jargon. By the end, viewers not only understand what chatbots do but how they think, making it a perfect foundation for anyone curious about conversational AI.
📫 Ask the Inbox
Q: “If a large language model just predicts the next word, how does it remember what I said earlier in a long chat?”
A: Think of an LLM as reading your message through a movable window of words called a context window.
Everything inside that window (usually the last few thousand words) gets turned into numbers the model can juggle at once. When the conversation runs longer than the window can hold, the oldest text slides out, much like older messages disappearing from the top of a phone screen.
To keep track of the broader thread, developers sometimes feed the model short “reminder” summaries of earlier points, so it still feels like the bot remembers the whole story without overloading its short‑term memory.
So, if a reply seems to forget details you mentioned way back, it’s not trying to be rude, its context window has simply moved on. Keeping prompts concise or recapping key facts along the way helps the model stay on track.
📌 Key takeaway: An LLM only “remembers” the most recent chunk of text it can fit in its context window, so short prompts and brief reminders keep long chats coherent.
🤔 Got Questions?
Have a question you want answered? Email [email protected] and you just might be featured in an upcoming issue!
This newsletter is your starter kit for mastering AI with confidence. We keep things simple, show real examples, and focus on quick wins you can repeat.
Stay curious, stick with it, and watch your skills grow week by week!
Until next time :)