AI with Kyle - Daily AI News and Updates

AI with Kyle - Daily AI News and Updates

Download on the App Store

AI with Kyle - Daily AI News and Updates episodes

  • 164: How to start VIBE CODING for free (no coding experience required, at all)

    Get the free Vibe Coding guide here: https://aiwithkyle.com/mini/5-builds

    Get AI-Ready with Kyle’s 5-Day Challenge: https://aiwithkyle.com/join

    Kyle’s AI Canon — Curated list of AI learning essentials: https://aiwithkyle.com/ai-canon

    Subscribe and turn on notifications to catch the next live stream: https://www.youtube.com/channel/UChlLglbHDASnoGkbjDeHnQg

    Summary: Today I stripped vibe coding back to the absolute basics. If you've been hearing the term thrown around and feeling like you need to be technical to get involved, you don't. I walked through exactly how to get started using Google AI Studio's hidden Build mode, which is completely free, requires zero setup, and lets you build working apps just by describing what you want in plain English. I built a social media growth projector live on stream, from nothing to a functioning tool with charts and milestone calculations, in about two minutes. The whole point was to remove every barrier I could think of. No paid tools, no terminal commands, no deployment headaches. Just open a browser, click Build, and start talking. I also put together a second tool, a task automation analyser, to show how quickly you can go from idea to something genuinely useful. If you've had projects sitting on a to-do list for months because you assumed you'd need to learn to code first, this is your way in. Start with a toy project, get the bug, and worry about the complicated stuff later.

    —— Time Stamps ——

    0:00 Getting Started with Vibe Coding for Complete Beginners

    2:41 Why Google AI Studio Build Mode is the Best Free Tool
    3:38 How to Use the Chat Interface and Voice Input
    4:15 Start with a Toy Project Not Your Dream App
    5:40 Live Build: Social Media Growth Calculator with Gemini
    10:00 Publishing and Deploying Your First App for Free
    11:06 Why Learning to Code Still Matters with AI
    11:44 Bugs vs Features: The Vibe Coding Feedback Loop
    14:28 How to Fix Bugs Using Screenshots and Error Logs
    15:28 From Toy Project to Real Business Tool Using Claude
    16:36 Planning Your Project Before Building
    19:53 Dropping a Project Plan into Google AI Studio Build
    20:47 What Beginners Should Skip: Auth, Databases and Deployment
    22:51 Vibe Five: 5 Mini Projects to Build Your Coding Portfolio
    24:55 Live Demo: AI-Powered "What Should I Automate First" Tool
    28:28 AI-Generated Automation Suggestions That Actually Work
    29:10 Recap: How to Start Vibe Coding Today
    31:28 Join the Community and Free Resources

    —— More Useful Resources ——

    Find everything else at https://aiwithkyle.com/

    33 min
  • 163: LEVEL UP: The AI Skill Tree (Most People Stuck at Level 1)

    Get AI-Ready with Kyle’s 5-Day Challenge: https://aiwithkyle.com/join

    Kyle’s AI Canon — Curated list of AI learning essentials: https://aiwithkyle.com/ai-canon

    Subscribe and turn on notifications to catch the next live stream: https://www.youtube.com/channel/UChlLglbHDASnoGkbjDeHnQg

    Summary: I've been promising this for weeks and finally got round to it. Tens of thousands of people asked for a proper guide to levelling up with AI. So today I built it. Seven levels of AI mastery, from "I'm basically using ChatGPT as Google" all the way through to running your own local models. I've mapped out the full progression path, complete with an RPG-style skill tree because apparently that's the kind of person I am. The core idea is simple: there's a logical order to learning this stuff, and most people are skipping straight to the fancy bits without the foundations. I walked through each level on the stream, from thinking with AI instead of just asking it questions, to setting up projects, personalising your setup with skills and custom instructions, connecting AI to your existing tools via MCP, vibe coding your own apps, building automations that work while you sleep, and finally going full AI operator. I also showed off a mini "five builds in five days" vibe coding curriculum I've been putting together, and built a stamp duty calculator live on stream just to prove the point. The whole thing will be fleshed out properly over the coming weeks, but the skeleton is there and I think it fills a gap that nobody else is really addressing.

    —— Time Stamps ——

    0:00 Introduction to the AI Native Skill Tree

    2:10 The Problem: Where Do I Start with AI?
    4:20 Level 1: Going Beyond Google
    5:12 Why Context and Prompting Matter
    7:48 Your AI Tech Stack in 2026
    9:09 Level 2: Building Persistent AI Projects
    10:50 Level 3: Personal AI, Custom Instructions & Skills
    14:03 Level 4: Connected AI & MCP Explained
    16:28 Level 5: The Vibe Coding Progression Path
    19:42 Level 6: AI Automation & Autonomous Agents
    22:25 Level 7: Becoming an AI Operator
    23:35 Why Existing AI Training Courses Are Broken
    25:56 All 7 Levels Recap & Mindset Shifts
    29:01 Best Vibe Coding Tools for Beginners
    31:17 Hosting Costs & Deploying AI Web Apps
    32:36 How to Find Your First Vibe Coding Project
    33:43 5 Builds in 5 Days: Vibe Coding Mini Course Preview
    38:14 Outro & How to Stay Updated

    —— More Useful Resources ——

    Find everything else at https://aiwithkyle.com/

    40 min
  • 162: 143M jobs scored by how at RISK they are from AI

    Get AI-Ready with Kyle’s 5-Day Challenge: https://aiwithkyle.com/join

    Kyle’s AI Canon — Curated list of AI learning essentials: https://aiwithkyle.com/ai-canon

    Subscribe and turn on notifications to catch the next live stream: https://www.youtube.com/channel/UChlLglbHDASnoGkbjDeHnQg

    Summary: Andrej Karpathy, co-creator of ChatGPT, quietly built a research agent, fed it the entire US job market and scored 143 million jobs on their AI exposure. The results went viral, then he deleted the whole thing. Thankfully someone cloned the repo before it vanished. The takeaway is blunt: if your job lives on a screen, you're looking at a 7, 8, or 9 out of 10 on the AI risk scale. Software developers, lawyers, customer support, travel agents, all deep in the red. Nurses, plumbers, electricians? Barely a scratch. We walked through the whole chart live and I'll be honest, it's equal parts fascinating and uncomfortable.

    Beyond the jobs data, we covered a genuinely brilliant story out of Australia where a bloke with zero medical training used ChatGPT and DeepMind's AlphaFold to design a personalised cancer vaccine for his dog. It worked. The tumour halved. The red tape took longer than the actual science, which tells you everything about where we are right now. We also talked about Meta quietly falling out of the AI race after their new model flopped internal benchmarks, and I walked through a bunch of Claude Code skill repositories so you can start plugging domain expertise straight into your AI workflow.

    —— Time Stamps ——

    0:00 Intro

    0:30 Karpathy's AI Job Exposure Chart: 143 Million US Jobs Scored
    5:01 Which Jobs Are Most at Risk from AI?
    8:37 Are Lawyers Really Safe from AI?
    9:39 Safest Jobs in the AI Age: Trades and Healthcare
    10:52 Live Q&A: Finding Your Job on the Chart
    13:44 Will AI Replace Teachers?
    15:29 The Trades Oversupply Trap
    16:41 Why You Need Multiple Income Sources Now
    19:50 Man Uses ChatGPT to Create a Cancer Vaccine for His Dog
    29:33 Meta Falls Behind: The AI Race Is Down to Three
    34:18 Best Free AI Documentaries on YouTube
    36:52 Claude Skills Explained: Expert Knowledge for Your AI
    39:13 Where to Find 88,000+ Claude Code Skills
    43:51 Claude 2x Off-Peak Usage Promotion Explained
    46:57 Wrap Up & Newsletter

    —— More Useful Resources ——

    Find everything else at https://aiwithkyle.com/

    49 min
  • 161: How to build interactive charts and diagrams with AI

    Get AI-Ready with Kyle’s 5-Day Challenge: https://aiwithkyle.com/join

    Kyle’s AI Canon — Curated list of AI learning essentials: https://aiwithkyle.com/ai-canon

    Subscribe and turn on notifications to catch the next live stream: https://www.youtube.com/channel/UChlLglbHDASnoGkbjDeHnQg

    Summary: Anthropic dropped the Visualizer for Claude today and I've spent all morning playing with it. The short version: you can now get Claude to create interactive charts, explainers and data visualisations directly in your conversation. The trick nobody's telling you is that you need to say "show me" to trigger it. I tested it on everything from compound interest explainers to debunking that nonsense going around TikTok about AI using more water than video streaming (spoiler: TikTok uses about 60 times more water per hour than ChatGPT, and that's being generous). You can even publish your visualisations as shareable artifacts, which is genuinely useful if you need to send something to a colleague or client without making them sign up for anything.

    The real story though is what this means practically. I used it this morning with my business partner to analyse our welcome email sequences, pulled in sales data across four different variants, and had Claude turn the whole thing into a visual report and then a slide deck. That process would have eaten an entire day six months ago. If you're running any kind of agency or reporting workflow, this is worth your attention. It's not going to destroy industries overnight, despite what some people are breathlessly claiming, but it's a proper useful addition to the toolkit. Available on all plans including free, which is a nice touch.

    —— Time Stamps ——

    0:00 Claude's New Visualizer Feature: What It Is & Why It Matters

    1:45 How to Trigger Claude's Visualizer: Use "Show Me"
    3:16 Debunking the AI Water Usage Myth: ChatGPT vs TikTok
    7:20 NotebookLM vs Claude Visualizer: Which Is Better for Learning?
    8:07 Interactive LLM Chunking Walkthrough: Learn AI Concepts Visually
    12:04 Publishing Claude Artifacts to the Web + RAG Pipeline Demo
    15:32 Visualizer API Access, Embedding Artifacts & Claude Plan Pricing
    18:18 Real Business Use Case: Analyzing Email A/B Sequences with Claude
    22:38 AI for Agencies: Auto-Generate Client Reports & Slide Decks in Minutes
    23:33 Why Most People Think AI Is Useless (The Free Tools Trap)
    26:03 Claude vs ChatGPT: Honest Comparison, Memory & Usage Limits
    30:16 Q&A: ChatGPT Memory Import, LLM Self-Hosting & Claude Code vs Cursor
    33:08 Q&A: AI Regulation, AGI Definition & Best Open Source LLMs to Self-Host
    36:51 Wrap-Up: Key Takeaways, Newsletter & Free Community Resources

    —— More Useful Resources ——

    Find everything else at https://aiwithkyle.com/

    38 min
  • 160: How to Get Started With Claude Code (NO Coding Experience)

    Get AI-Ready with Kyle’s 5-Day Challenge: https://aiwithkyle.com/join

    Kyle’s AI Canon — Curated list of AI learning essentials: https://aiwithkyle.com/ai-canon

    Subscribe and turn on notifications to catch the next live stream: https://www.youtube.com/channel/UChlLglbHDASnoGkbjDeHnQg

    This video is from the ‘AI with Kyle Live Stream’

    First aired: 12 March 2026

    Summary: Finally did it. I've been putting off this Claude Code explainer for weeks, partly because simplifying a genuinely complex topic is tricky, and partly because I was just a bit intimidated by it. But here we are. This is a proper beginner's guide to getting started with Claude Code if you're not a developer, have never touched a terminal, and have no idea what a GitHub repo is. We cover the three main ways to access Claude Code (web, desktop app, and IDE), why the desktop app is the one to go for, and how to connect it to GitHub without losing your mind.

    The bigger point here isn't really about the tool itself. It's about what becomes possible when you stop assuming software is someone else's job. Claude Code lets you build small, genuinely useful things: personal micro-tools, scrapers, dashboards, automations. We built a (very rough) Space Invaders clone live, just to prove the point. I also walked through Gary Tan's GST Stack setup, which turns Claude Code into something resembling a full engineering team. It's early days for most people on this, and that's fine. Starting small is exactly the right move.

    —— Time Stamps ——

    0:00 Claude Code for Non-Coders: What It Is and Why You Need It

    4:35 Why the Official Docs Look Scary (And How to Skip the Hard Stuff)
    6:10 3 Ways to Access Claude Code: Web App, Desktop & VS Code
    9:26 Setting Up the Claude Desktop App & Understanding GitHub
    13:36 Live Demo: Claude Code Builds a Space Invaders Game From Scratch
    16:35 Previewing Your Build: Local Deployment Without the Terminal
    19:13 What Comes Next: Project Structure, Deployment & Adding a Backend
    22:12 Clone Any Claude Code Setup: Installing Gary Tan's GStack Live
    26:47 Q&A: Claude Code Pricing — $20, $100 or $200 Per Month?
    28:20 Free Alternatives: ChatGPT Codex, Anti-Gravity & the CLI Option
    31:41 Should You Still Learn to Code? The Honest Answer for 2025
    34:03 AI Upskilling & Junior Dev Jobs: Who's Really at Risk?
    35:43 The Military AI Truth: All AI Companies Are Taking Pentagon Contracts
    37:52 How to Make $3K Per Hour Teaching AI (No PhD Required)
    39:41 Raspberry Pi + Claude Code: Hardware Projects & Pi-Hole Setup
    40:45 Can Claude Code Build iOS & Android Apps? Xcode SDK Integration
    42:49 Anthropic's Supply Chain Risk: Why Amazon Said Nothing
    44:31 Kyle's Raspberry Pi Badge Project & Getting Started with Hardware
    46:01 Context Window Limits: How to Handle a 1,000-Page Document in Claude Code
    47:32 Wrap-Up: Skills, MCP & Advanced Claude Code Sessions Coming Next

    —— More Useful Resources ——

    Find everything else at https://aiwithkyle.com/

    50 min
  • 159: Complete Guide to Moving from ChatGPT to Claude (And KEEP Your Memories)

    Get AI-Ready with Kyle’s 5-Day Challenge: https://aiwithkyle.com/join

    Kyle’s AI Canon — Curated list of AI learning essentials: https://aiwithkyle.com/ai-canon

    Subscribe and turn on notifications to catch the next live stream: https://www.youtube.com/channel/UChlLglbHDASnoGkbjDeHnQg

    This video is from the ‘AI with Kyle Live Stream’

    First aired: 11 March 2026

    Summary: A lot of people have been making the jump from ChatGPT to Claude lately, for political reasons, for professional ones, or simply because they're tired of an AI that agrees with absolutely everything they say. The transition isn't as smooth as you'd hope though, and if you arrive expecting the same experience you had on ChatGPT, you're in for a few surprises — usage limits, no image generation, and a very different philosophy baked into the product.

    In this one I walk through everything you actually need to know before making the switch: the model tiers and what you can realistically use on a $20 plan, how Projects and Memory work, the connectors, Skills, and where Claude Code and Cowork fit into the picture. Claude is built for professionals who want straight answers rather than flattery, and once you understand how it thinks, it's a genuinely better tool for serious work. It just takes a little getting used to.

    —— Time Stamps ——

    0:00 Moving from ChatGPT to Claude: What the TechRadar Article Got Right

    2:05 The Painful Truth About Claude's Usage Limits and Pricing
    3:00 Why Claude Is Less Sycophantic Than ChatGPT (And Why That Matters for Work)
    6:47 What Claude Can't Do: No Images, Video or App Store
    8:54 The Three Claude Models Explained: Opus, Sonnet and Haiku
    13:45 How to Import Your ChatGPT Memory into Claude
    17:01 Setting Up Claude: Custom Instructions, Model Selection and Extended Thinking
    20:10 Artifacts Explained: Claude's Built-In App and Document Builder
    21:15 Connectors and Skills: Linking Claude to Your Tools and Workflow
    25:21 Chat, Code and Cowork: Understanding the Full Claude Ecosystem
    27:09 Claude's Biggest Problem Right Now: No Unified AI Brain Across Tools
    29:35 Q&A: ChatGPT's Manipulative Teasing Cadence Is Ad Preparation
    32:00 How Claude Code Is Killing Paid Apps: The LinkTree Example
    36:52 Anthropic's AI Job Displacement Chart: Separating Reality from Hype
    43:13 Claude Code Adds Code Reviews: Is Code Rabbit Dead?
    46:47 Vibe Coding for Everyone: The $30M Teen App and What It Means
    49:51 The New York Times Blind Writing Test: Can Claude Beat Carl Sagan?
    55:10 Wrap Up and What's Next: Claude Code Deep Dive Coming

    —— More Useful Resources ——

    Find everything else at https://aiwithkyle.com/

    57 min
  • 158: Complete Claude Cowork Setup Guide (Automating Your Business in 2026)

    Get the Free Cowork Setup Guide: https://aiwithkyle.com/mini/claude-cowork

    Get AI-Ready with Kyle’s 5-Day Challenge: https://aiwithkyle.com/join

    Kyle’s AI Canon — Curated list of AI learning essentials: https://aiwithkyle.com/ai-canon

    Subscribe and turn on notifications to catch the next live stream: https://www.youtube.com/channel/UChlLglbHDASnoGkbjDeHnQg

    This video is from the ‘AI with Kyle Live Stream’

    First aired: 9 March 2026

    Summary: We are at a genuinely interesting moment with AI — and most people are still sleeping on it. In this session I walk through Claude Cowork, what it actually is, why it's meaningfully different from chatting with an AI, and how I'm using it to automate tasks that would otherwise eat my day. We're talking TikTok reply bots, ad audits, scheduled workflows — all running in the background while I get on with other things.

    This isn't about prompting AI anymore. It's about setting up systems once and letting them run. I cover the full Claude ecosystem, how skills work, what MCP connectors are, and the honest caveats too — cost, desktop-only limitations, token burn. If you've been curious about where AI is actually headed in 2026, this is a decent place to start.

    —— Time Stamps ——

    0:00 Welcome to the Claude Ecosystem

    0:53 $830 Billion Stock Wipeout & The 4 Big AI Shifts
    3:43 The Big Shift: AI That Does the Work, Not Just Talks
    6:00 Getting Started: Download, Cost & Setup Requirements
    7:00 How Cowork Works: Files, Folders & the Employee Model
    7:37 Live Demo: Automating TikTok DM Replies
    9:50 From Prompts to Files: The New Way to Delegate to AI
    14:20 Live Demo: Invoice Automation Built from Scratch
    20:22 Claude Skills: Plug In Any Expertise Instantly
    22:07 Live Demo: Automated Meta Ads + Stripe Daily Report
    23:34 One Person. Ten People's Output. Here's Why.
    24:15 Cowork Caveats: Cost, Desktop-Only & What Can Go Wrong
    29:04 The Full Ecosystem: Chat vs Cowork vs Code vs MCP
    31:06 Q&A: Claude for Business, GPT 5.1 Retiring & A Student Goes Viral
    35:39 Q&A: Human in the Loop, AI Adoption & Building a Daily Report
    40:08 Q&A: What I Automate on TikTok (and What I Don't)
    40:38 Q&A: AI Royalties, App Costs & the $30M Cal AI Story
    42:43 Agent Wars: Claude vs ChatGPT Codex vs Google Gemini
    45:31 Anthropic's 33-Page Skills Guide & Building Custom Skills
    47:21 Q&A: Switching from ChatGPT, Security, WordPress & AI Careers
    57:45 Wrap-Up & Get the Free PDF Guide

    —— More Useful Resources ——

    Find everything else at https://aiwithkyle.com/

    1 hr
  • 157: Exact AI System I Use to Create 20+ Content Pieces A DAY

    Free guide - Turn 1 recording into 20+ pieces of content: https://aiwithkyle.com/mini/content-engine

    Get AI-Ready with Kyle’s 5-Day Challenge: https://aiwithkyle.com/join

    Kyle’s AI Canon — Curated list of AI learning essentials: https://aiwithkyle.com/ai-canon

    Subscribe and turn on notifications to catch the next live stream: https://www.youtube.com/channel/UChlLglbHDASnoGkbjDeHnQg

    This video is from the ‘AI with Kyle Live Stream’

    First aired: 6 March 2026

    Summary: Identified and restructured em dashes throughout SEO summary

    I spent this session pulling back the curtain on the exact content workflow I use to run my business off the back of a 300,000-strong audience. The honest version: I do a one-hour live stream every day, record 15 short videos afterwards, push out a newsletter, and convert bits of it into blog articles. That sounds unhinged, and it probably is. But the point wasn't to make you copy me. It was to show you how a tiny team can operate like a full media company when you use AI to handle the repurposing rather than the creating.

    The key thing I wanted to get across is that AI slop is already everywhere and people can smell it. LinkedIn is basically a skip fire at this point. The way to beat that is to keep yourself in the loop. Record your actual voice, your actual opinions, your actual hot takes, and then let AI tidy, reformat, and multiply that into newsletters, short-form scripts, and blog articles. Fifteen minutes of rambling into your phone can become 20+ pieces of content. Distribution is the game now, not building. This workflow is how you play it.

    —— Time Stamps ——

    0:00 How I Use AI to Create 30+ Content Pieces Daily

    1:28 The One Hour Content Engine Explained
    2:31 Why Distribution Beats Product Building in 2026
    3:39 The AI Slop Problem on LinkedIn and Social Media
    5:02 Human in the Loop: The Secret to Authentic AI Content
    6:29 My Complete Daily Content Workflow Overview
    8:01 Stage 1: Setting Up Your AI Research Agent
    10:20 Stage 2: Why Recording Your Raw Thoughts Is Everything
    12:10 Stage 3: Turning Transcripts Into a Daily Newsletter
    14:12 Stage 4: AI-Generated Short Form Video Scripts
    16:09 Why TikTok and Instagram Algorithms Reward Yapping
    16:49 How Content Stacking Creates a Flywheel Effect
    17:40 Stage 5: Turning Short Videos Into SEO Blog Articles
    19:25 Building a Content-to-Sales Funnel That Works
    20:26 Best Newsletter Platforms: Beehive vs Substack vs Kit
    21:32 Your Week 1 Action Plan to Start Today
    23:47 Q&A: Why NotebookLM Is the Best Free AI Tool
    25:22 How I Got Into AI Training as a Humanities Grad
    26:24 Overcoming Public Speaking Fear: My Roadmap
    27:00 AI Generated vs AI Assisted Content: Where's the Line?
    28:47 The Hormozi Quote That Changed My Content Strategy
    30:23 Closing and How to Get the Free PDF Guide

    —— More Useful Resources ——

    Find everything else at https://aiwithkyle.com/

    31 min
  • 156: AI Update: Anthropic Pentagon Fallout, Google Workspace CLI & NotebookLM Goes Cinematic

    Get AI-Ready with Kyle’s 5-Day Challenge: https://aiwithkyle.com/join

    Kyle’s AI Canon — Curated list of AI learning essentials: https://aiwithkyle.com/ai-canon

    Subscribe and turn on notifications to catch the next live stream: https://www.youtube.com/channel/UChlLglbHDASnoGkbjDeHnQg

    This video is from the ‘AI with Kyle Live Stream’

    First aired: 5 March 2026

    Summary: Anthropic had quite the week. Kicked out of the Pentagon, labeled a "supply chain risk" (a designation usually reserved for Chinese state-linked companies), only for Dario Amodei to be back at the negotiating table days later. Sam Altman, meanwhile, swooped in suspiciously fast to fill the gap, which didn't go unnoticed. Over a million and a half people quietly migrated to Claude, pushing it to number one on the App Store for the first time in its life. The irony being, of course, that Anthropic are perfectly happy building lethal autonomous weapons. They just want Americans kept out of the blast radius.

    On the quieter but arguably more interesting side of things, Google dropped a native command line interface for the entire Workspace ecosystem, which is the kind of boring plumbing update that builders absolutely lose their minds over, and rightly so. AI agents can now talk directly to Gmail, Drive, Calendar, Sheets and more without the usual browser automation nonsense. NotebookLM also got a cinematic video upgrade, turning static slideshows into full motion AI-generated explainers. It's been a dense week.

    —— Time Stamps ——

    00:00 Anthropic vs Pentagon: The Deal That Could End a $60B Company

    11:34 Google Workspace CLI: The Biggest AI Release Nobody's Talking About
    17:08 NotebookLM Gets Cinematic AI Video (Google's Quietly Massive Update)
    22:24 Claude vs ChatGPT: The Honest Pricing Breakdown
    27:44 How to Make $3,000/Hour Teaching AI to Businesses
    30:33 Free Vibe Coding Tools: AntiGravity vs Google AI Studio
    34:31 OpenAI Model Deprecations: What's Being Deleted & When
    36:13 Q&A: How to Actually Get Started with Claude Skills
    38:37 The MIT AI ROI Study Everyone Misread
    41:29 UK's £40M AI Lab: Why It's Already Too Late
    43:42 Final Thoughts & Webinar Details

    —— More Useful Resources ——

    Find everything else at https://aiwithkyle.com/

    45 min
  • 155: The Agentic Era Is Officially Here & GPT 5.3 “less cringe” Drops

    Get AI-Ready with Kyle’s 5-Day Challenge: https://aiwithkyle.com/join

    Learn how to become an AI trainer - Live Webinar: https://aiwithkyle.com/ai-workshop-series/webinar

    Kyle’s AI Canon — Curated list of AI learning essentials: https://aiwithkyle.com/ai-canon

    Subscribe and turn on notifications to catch the next live stream: https://www.youtube.com/channel/UChlLglbHDASnoGkbjDeHnQg

    This video is from the ‘AI with Kyle Live Stream’

    First aired: 4 March 2026

    Summary: GPT dropped 5.3 today, marketed brilliantly as "more accurate, less cringe." That tells you everything about how badly 5.2 landed. Nobody liked it, it gaslit users into thinking they needed to calm down before asking about banana bread, and OpenAI know it. There's also a 5.4 reasoning model apparently dropping this week, though at this point it feels less like a product launch and more like a fire extinguisher. Cancellations are up 400% and people are migrating to Claude in droves.

    The bigger conversation though is the one that actually matters: we've crossed a genuine threshold in AI. December 2025 was the quiet inflection point where agents stopped being a buzzword and started actually working. For three years, AI was something you talked to. Now it's something that does the work for you, handling coding, bookkeeping, the lot. Andrej Karpathy put it well: before December, coding agents basically didn't work. Since December, they basically do. That's not a gradual curve, that's a hard line. If you haven't started playing with agentic tools yet, now's a very good time to start paying attention.

    —— Time Stamps ——

    0:00 GPT-5.3 "Less Cringe" Update: What OpenAI Actually Changed

    3:23 GPT-5.4 Teaser: Is a Reasoning Model Dropping This Week?
    7:12 The 4 Big AI Shifts That Changed Everything (Ethan Mollick)
    10:13 From Chatting to Doing: The Real Agentic AI Revolution
    19:27 Live Demo: Claude Cowork Handles My Entire Bookkeeping
    21:20 White Collar Jobs at Risk Since December 2025
    22:08 Q&A: Best Chinese & EU AI Alternatives (Qwen, Kimi, DeepSeek)
    24:47 Will China Keep Open-Sourcing AI? The Geopolitics Explained
    29:18 Switching to Claude: Why Usage Limits Are Shocking New Users
    33:47 AI Safety Horror Story: Agent Deleted an Entire Email Inbox
    38:08 Anthropic vs Pentagon: The Nuance Everyone Is Getting Wrong
    43:04 WTF Happened in 2025: The December AI Inflection Point
    45:35 OpenClaw's Explosive GitHub Growth: The Numbers Are Staggering
    46:57 Claude Code Writes Itself: AI Development Going Vertical
    48:26 How to Get Started with Agentic AI: Cowork, OpenClaw & More
    54:00 Is Claude's $100/Month Plan Worth It? Real Cost Breakdown
    56:41 Outro: Newsletter, YouTube & Free Webinar

    —— More Useful Resources ——

    Find everything else at https://aiwithkyle.com/

    59 min

About AI with Kyle - Daily AI News and Updates

From the publisher's feed

AI With Kyle - Daily AI News With Zero Hype, Zero BS AI With Kyle is the daily podcast for people who want to understand artificial intelligence without hype...