196: AI 101: Tokens, Context Windows & Why AI Makes Dumb Mistakes Explained

196: AI 101: Tokens, Context Windows & Why AI Makes Dumb Mistakes Explained

Get AI-Ready with Kyle’s 5-Day Challenge: https://aiwithkyle.com/join Subscribe and turn on notifications to catch the next live stream: https://www.youtube.com/channel/UChlLglbHDASnoGkbjDeHnQg Summary:  Tokenization is one of those concepts that sounds technical until you realise it explains basically everything. Why AI can't count the letters in "raspberry." Why a bigger context window sometimes makes things worse. Why your API bill might be quietly spiralling. This is a 101 lesson on tokens: what they actually are, how models use them, and why the common assumption that a token equals a word is close but not quite right. The second half gets into context windows, the working memory of a model, and why the "just give it everything" instinct is both expensive and counterproductive. Whether you're building on the API or just a heavy user wondering why long threads start getting weird, understanding this stuff changes how you work. Also, token maxing is a trend and it is, in my view, a bit dumb. We get into that too. —— Time Stamps —— 0:00 Intro: What Tokenization Is & Why It Matters 1:27 Tiktokenizer Demo: How Words Break Into Tokens 4:18 Token IDs: How Models Store Words as Numbers 5:35 Tokens Across Languages: Why English Is Cheapest 7:59 The Raspberry Problem: Why AI Can't Count Letters 9:49 Context Windows Explained: What "1 Million Tokens" Means 11:40 Context Compaction: Why Chats No Longer Cut You Off 13:27 Why a Bigger Context Window Isn't Always Better 16:18 API Pricing: How Tokens Drive Your AI Costs 18:42 "Token Maxing": The Wasteful AI Trend to Avoid 20:13 Finding Real Context Limits & DeepSeek vs GPT Pricing 22:36 Rules of Thumb: Optimize for Useful Context 25:12 Wrap-Up & Andrej Karpathy Recommendation 26:20 Newsletter & Sign-Off — Useful Resources —— Find everything else at https://aiwithkyle.com/

195: How to Use Codex Pulses (and Delete Your To-Do App)

Get AI-Ready with Kyle’s 5-Day Challenge: https://aiwithkyle.com/join Subscribe and turn on notifications to catch the next live stream: https://www.youtube.com/channel/UChlLglbHDASnoGkbjDeHnQg Summary:  Codex Pulses are dedicated AI threads that run on a schedule, collect everything you throw at them, and send you a triage report every morning. Credit to Dan Shipper for the original idea. I tried it, built my own version, and haven't opened a to-do app since. This is the full setup walkthrough. The system works like this: you speak into the Codex mobile app on your phone, it syncs to your home computer, and your pulses do the processing overnight. Thought capture, to-dos, things to buy, content ideas. Each gets its own thread, its own rules, its own output. No coding required, no complex configuration (like Open Claw), just chat. —— Time Stamps —— 0:00 Intro: What Are Codex Pulses? 0:56 Dan Shipper's Codex Power User Method 2:07 What Is a Pulse? Building Your Own Threads 3:29 Codex vs OpenClaw: Why Focused Beats Broad 4:02 Capturing on Mobile: Your Home Computer as Command Center 5:41 Daily Reviews & Proactive AI Automation 7:32 Kyle's 5 Capture Pulses: Thoughts, To-Dos & Content Ideas 10:37 To-Do & Buy Capture: Auto-Sorting Your Purchases 12:05 The Codex Mobile App: The Real Game-Changer 13:20 Getting Started: Codex Pricing & Setup Walkthrough 15:15 Recap & Wrap-Up — Useful Resources —— Find everything else at https://aiwithkyle.com/

194: The AI Job Audit: Will AI Take Your Job?

Get AI-Ready with Kyle’s 5-Day Challenge: https://aiwithkyle.com/join Subscribe and turn on notifications to catch the next live stream: https://www.youtube.com/channel/UChlLglbHDASnoGkbjDeHnQg Summary: AI isn't going to tap you on the shoulder and tell you you're redundant. It's quieter than that. What's actually happening is a slow compression of the tasks that make up your working week, and if you're not paying attention to which ones are at risk, you'll be caught out. In this video, I walk you through how to break your job down into its component tasks, identify which ones are genuinely defensible, and work out where to focus your energy. I've also built a diagnostic tool that does a lot of the heavy lifting for you. You put in your job title, add a description, and it maps your weekly tasks against labour market exposure data to give you a real picture of where you stand. It's free. The link's above. But even if you want to run this manually first, I show you exactly how to do that in this video too. It's not about doom, it's about knowing where you actually are. —— Time Stamps —— 0:00 Will AI Take Your Job? The Real Threat Explained 1:30 Why Every Job Is a Bundle of Tasks 2:46 Exposed Edges vs Defensible Tasks (Accountant & Zookeeper) 4:06 How AI Quietly Shrinks Headcount 6:04 Why Gen Z and Entry-Level Workers Get Hit First 7:32 The Productivity Myth and the Demand Problem 9:27 Second-Order Effects: Why No Industry Is Safe 10:54 The DIY Job Audit: Map Your Weekly Tasks 12:51 Build Your Human Edge: Skills AI Can't Replace 14:43 Your Action Plan and the AI Job-Risk Tool — Useful Resources —— Find everything else at https://aiwithkyle.com/

193: Here's how I find what to automate with AI (and what NOT to)

Get AI-Ready with Kyle’s 5-Day Challenge: https://aiwithkyle.com/join Subscribe and turn on notifications to catch the next live stream: https://www.youtube.com/channel/UChlLglbHDASnoGkbjDeHnQg Summary: I've seen too many people rush straight into building AI agents for tasks that don't need them. This is the corrective. Starting from a viral prompt retweeted by Greg Brockman, OpenAI's co-founder, I walk through a proper workflow audit - how to find the tasks in your actual working week that are genuinely worth automating, and how to spot the ones that aren't. The framework is simple: audit your real work first, then match the fix to the job. A reusable prompt in a Word doc beats a managed cloud agent if that's all the task needs. I also show you how to put your AI into interview mode to surface repeating workflows you didn't even know you had. A practical system for getting more done with less friction. —— Time Stamps —— 0:00 The Viral Self-Improvement Prompt (Greg Brockman + Vaibhav) 4:06 Why You Shouldn't Automate for the Sake of It 5:29 How to Audit Your Real Weekly Workflows 7:19 When NOT to Automate: Avoid the Over-Engineering Trap 9:17 Put ChatGPT in Interview Mode (Live Demo) 13:04 The Simplest Tools First: Prompts, Checklists, and Templates 15:01 Stepping Up: Zapier, Make.com, and n8n Automations 16:10 Managed Agents, Sub-Agents, Scripts, and Custom Apps 17:18 Fit The Fix To The Job: Don't Get Blinded by Viral AI 19:10 Wrap Up + Where to Get the Prompt — Useful Resources —— Find everything else at https://aiwithkyle.com/

192: Google I/O: What You Actually Need to Know

Get AI-Ready with Kyle’s 5-Day Challenge: https://aiwithkyle.com/join Subscribe and turn on notifications to catch the next live stream: https://www.youtube.com/channel/UChlLglbHDASnoGkbjDeHnQg Summary: Google I/O is over, the dust has settled, and the takes have mostly been garbage. So here's what actually matters from Google's big conference this week. Hint: it's not the video generator. The story everyone missed is Google's move to embed agentic AI directly into Search. That's where 70% of their revenue comes from, and they're knowingly blowing it up. I walk through why that's a genuinely brave decision, what it means for SaaS businesses, how it compares to the OpenAI approach, and why distribution will always beat product quality. Also: Gemini 3.5 Flash benchmarks, Antigravity 2's messy launch, and why I'm not particularly impressed with any of it,  but still think this was one of the more important AI weeks of the year. —— Time Stamps —— 0:00 Intro: Cutting Through the Google I/O Hype 1:10 Why Google Conferences Look Flashy But Say Nothing 2:13 Distribution Beats Product: The Slack vs Microsoft Teams Lesson 3:53 Google's Confusing AI Product Sprawl (Jules, Opal, Antigravity, Spark...) 5:53 Gemini 3.5 Flash: Benchmarks and Why People Aren't Impressed 9:07 Google's Hidden Edge: Bringing AI Directly to Users 10:17 The Real I/O Story Nobody Covered: AI Inside Google Search 12:16 Why Google Must Cannibalize Their $82 Billion Ad Empire 14:46 How Google Agents Will Demolish Lightweight SaaS Companies 17:07 Project Spark: Google's Jarvis-Style AI Assistant 18:21 Antigravity 2: The Codex Clone That Flopped 20:21 Final Verdict: What Actually Matters from Google I/O — Useful Resources —— Find everything else at https://aiwithkyle.com/