Matt Maher - Videos
Back to Channel
I Asked GPT-5 to Rebuild My 100K Line App. Its First Move Changed Everything.
I expected it to start coding. Instead, it stopped, studied the entire application, and made a decision I've only ever seen experienced architects make. Six hours later, 150 files had changed — and...
GPT-5.4 Got the Best Score I've Ever Seen — Then I Found Something Stranger
GPT-5.4 scored 95% on my planning benchmark — the highest I've ever recorded. But while I was testing it across every tool I use, a pattern showed up in the data that I genuinely did not expect. An...
My Biggest AI Unlock — It Does Everything
It takes 30 seconds to set up. It sounds like absolutely nothing. And it replaced every other AI tool I use. I spent a week trying to figure out how to explain this because every time I described ...
Why Your AI-Improved Code Feels Wrong
AI coding tools are making your software functionally better — but something feels off. Here's what's actually happening. Every optimization makes it better. Every optimization scrapes off a littl...
Claude Code Remote Control Is About to Change Things
There's a new Claude Code feature, and it's the first thing that actually solved a problem I didn't realize I was still carrying around. Your sessions don't stop when you walk away. Now you can see...
Gemini 3.1 Pro Faltered — And Revealed Something Bigger
I've been running the same planning benchmark against frontier models for six months. Every one of them clusters between 70-90%. Gemini 3.1 Pro scored so far outside that range, I had to test it fi...
The One Shift That Takes You to Level 3 With AI
I spent a year figuring out why AI worked brilliantly sometimes and felt completely useless other times. Turns out the answer had nothing to do with the tools. Most people using AI right now are s...
Something Is Happening Inside Claude Code Agent Teams
I designed every agent. I understand the architecture. And the output still surprised me — not because it was good, but because I can't trace how it got there. I built an ideation system using Cla...
You're Throwing Away Your Best Thinking Before AI Sees It
Every time you type a prompt, you edit it first. You compress your thought down to the short version — and the part that actually matters never makes it in. There's a fix, and every AI company on t...
Codex Spark's Real Value Isn't Coding — Here's What It Is
I spent an hour using Codex Spark before I read a single thing about it. That hour told me more than any benchmark chart could. OpenAI just shipped Codex Spark — a smaller variant of GPT-5.3 Codex...
60 Seconds to Build a Claude Code Skill That Lasts Forever
If you're prompting Claude Code from scratch every time, you're working way too hard. I was too — until I figured this out. Claude Code skills are the most underleveraged feature in AI right now. ...
Opus 4.6 vs Codex 5.3: Tested in the Real World
The benchmark charts said these two models were basically tied. So I stopped reading charts and ran three experiments instead. What I found had very little to do with which model is smarter. Opus ...
What the Codex App Really Controls (The Visibility Rule Explained)
OpenAI just shipped a standalone Codex desktop app for macOS. Not a ChatGPT tab. Not an IDE plugin. A standalone application sitting above your entire development environment. But there's a visibi...
AI Techniques Distilled From Thousands of Hours of Real Work
You can do everything right — best model, best prompts — and AI will still silently drop 40% of what you asked for. That's the failure mode nobody warns you about. These are the six practices that ...
Claude Code Makes Video Now (It's Not What You Think)
You've heard people say Claude Code can make video now. They're right — sort of. What it actually does is something most developers will find far more useful than AI-generated video, and it works t...
The Most Powerful Claude Code Pattern I've Found
I asked Claude a simple question. What came out was a full software development workflow. Two skills. Capture on one side, build on the other. Files in a folder, sub-agents doing the work, commits...
ChatGPT Is Burying Your Best Ideas #ClaudeCode
ChatGPT doesn't save your work. It buries it. Every outline, every deci...
Why ChatGPT Fails at Real Work
ChatGPT is incredible at conversation. It's terrible at work. There's a difference nobody talks about—and a fix that changes everything. I'll show you a dead-simple pattern: a folder, an AI that c...
AI Gave You Coding. Now It Gives You Design.
A year ago, you probably said the same thing about building apps. "I can't code." Then AI changed that. Design is next—and it's not about becoming an artist. This video hands you the vocabulary, t...
Can You Name What You're Looking For?
The skill that made you senior—reading every diff, knowing every pattern, being the arbiter of "good enough"—might be the thing that's slowing you down now. I've watched the curve change. A year a...
First Look: Claude Cowork Is the Real Deal
Anthropic just dropped something I've been waiting for: Claude Cowork. It's basically Claude Code without the terminal—all that agentic, go-do-anything-on-my-computer power, but with a proper inter...
I Built the Same App 9 Times—The Results Made No Sense
I thought GPT-5 Pro would crush the cheaper models. 15 minutes vs 15 seconds to generate a PRD—there had to be a difference. Then I built the same app 9 times and discovered something I wasn't expe...
Claude Code's Chrome Extension Does Way More Than You Think
Most people think this is just another AI browser sidebar. I thought the same thing until I saw what it could do with Claude Code. The Claude team quietly shipped a Chrome extension that does far ...
You're Not Crazy. AI Really Does Ignore You.
You tell it something important. It nods, gets to work, and five minutes later—does exactly what you told it not to do. Then it apologizes. Promises to do better. And five minutes after that? Same ...
Claude Code: This Skill Is No Longer Optional
This isn't about becoming a programmer. It's about developing the skill that will matter for everything. The gulf between people who understand these tools and people who don't is widening fast. A...
Vibe Coding: From Party Trick to Real Tool
Vibe coding isn't what you think it is. If your mental image is "one prompt, one toy app, neat trick"—that's the old version. That was just the doorway. What's on the other side is something else e...
Why GPT-5.2 Codex Didn't Impress Me
I ran GPT-5.2 Codex through the same challenge as six other models. Identical codebase. Identical prompts. Identical evaluators. The results surprised me—but not in the way OpenAI probably hoped. ...
2025 Was the Year AI Changed Software Engineering
From tab-tab-tab completion to building entire systems with a single prompt—2025 wasn't just a year of progress. It was a paradigm shift that laid the foundation for everything that comes next. Th...
ChatGPT vs Claude: One Sounds Like a Blog (Here's the Fix)
Same prompt. ChatGPT sounds like a blog post. Claude sounds like a person. These models are NOT interchangeable—and I built a free tool to prove it. This video breaks down what actually happens wh...
AGI Will Create Billionaires. Maybe Ten of Them.
AGI promises to cure cancer, revolutionize education, extend human life. But who actually gets access to those solutions? We already have a $2 million gene therapy that cures a deadly childhood di...
GPT-5.2 vs Opus 4.5: The Ultimate Coding Benchmark
A year's worth of code. Built in hours. GPT-5.2 vs Opus 4.5 on my hardest benchmark yet—and both crushed it. I designed this benchmark differently than anything you've seen before. No cherry-picke...
I Gave Claude Code a Different Job. It Changed Everything.
Claude Code isn't just a coding tool—I've been using it for months to build multi-agent workflows that have nothing to do with code. Meeting notes, research systems, image pipelines, even jingle ma...
How I Make Claude Code Cook
70% of my success happens before I write a single line of code. Here's the four-phase system I use to ship software with AI—Define, Refine, Multiply, Expand—and why most people skip the part that m...
Claude Code: 3 Unlocks to Level 20
I asked Claude Code to run my tests, merge a PR, and cut a release. It did all of it—no GitHub UI, no manual commands. Just me and a terminal shipping a feature to production. Plan Mode spawned fi...
Opus 4.5 Sees the Whole System
Sonnet 4.5 and Codex Max High both failed on a React bug. For hours. Death spirals, circular fixes, no progress. Then Opus 4.5 and Gemini 3 dropped—and solved it in one shot. But here's what got m...
I Tried to Build a Prompt Manager in One Hour—Chaos Ensued
I tried to build a fully working Prompt Manager in one hour… and everything went off the rails fast. But with parallel agents, chaotic debugging, and a real ticking clock, something surprising happ...
New Gemini 3 Antigravity - My New IDE
Google just dropped Gemini 3.0, and I spent days testing its engineering capabilities—here's what blew me away. In this deep dive, I explore Anti-gravity (Google's new VS Code clone), compare Gemin...
New GPT-5.1: 80% Token Reduction?
I ran the same architectural analysis task across all their models to see the real differences. The results surprised me: Codex Mini used 400,000 tokens vs 160,000 for Codex High, but finished in h...
Starting AI Engineering Today? This Is What I'd Do
IT IS NOT TOO LATE! If I was starting today as a traditional engineer, here's exactly what I'd do: 30 days, three web applications built with ChatGPT Codex or Claude Code, zero frameworks, one focu...
I Deployed Code from My Couch — Here’s How It Works
Today’s developer reality: code, cloud… and couch? In this video, I show how I can dream up a change to an app, make it, and push it live—all without leaving the couch. Using GitHub, Vercel, and C...
I Used Both Claude Code and Codex for Weeks — Here’s What Surprised Me
I spent weeks using both Claude Code and Codex side by side — not to crown a “winner,” but to figure out how each one actually thinks. If you’ve ever wondered why one tool just “clicks” with you a...
Claude Code’s Hidden Subagents: Plan + Explore → Team Mode
Claude Code just quietly got a team. Two hidden subagents — Plan and Explore — change how it thinks, builds, and collaborates. Claude Code now includes two new subagents that mark a real shift in ...
AI Tools Are Outpacing How We Build Software
A tiny “Apply” button reveals something bigger — our dev workflows aren’t built for AI speed. I’ll show how Codex Cloud exposes the gap between tools and process. The way we build software is star...
AI Ghost Stories: Is It True?
Ever wonder what happens when your AI keeps working… even after you turn it off? This Halloween, I’m sharing a chilling (and totally fun) short story about an AI assistant that wouldn’t stop posti...
The Real Reason AI Is Going Back to the Terminal
Why is everyone suddenly obsessed with CLIs again? From Claude Code, Gemini CLI, OpenAI’s Codex, and even Cursor, it feels like every major AI company is rushing back to the terminal. But… why? In...
The New Engineering Mindset for the AI Era
After two years of breakthroughs, models that code, reason, and write — why does it still feel like we’re falling behind? We’ve never had more powerful AI, yet we’ve never felt more overwhelmed tr...
ChatGPT Is Now a Browser?! Full Tour of OpenAI Atlas
OpenAI just launched its first browser — ChatGPT Atlas — and it’s already changing how we think about browsing with AI. In this video, I take you through Atlas’ new AI-integrated experience: from i...
Claude Code Web Is the Future of AI Development
Claude Code has finally arrived on the web — and it’s a game-changer for how we write and scale code. In this video, I walk through Claude Code on the Web, the new browser-based experience from Ant...
Claude Skills Review — Early, But Promising
Claude’s new Skill System might quietly be one of the most useful features Anthropic has added — even if it’s still a little early. In this video, I take you through what Skills actually are, how ...
I Taught an AI Workshop About Nothing—and It Worked
What happens when you try to teach a team of engineers nothing—and it actually works? In this workshop, I set out to create a team of new AI engineers ready to tackle any problem, using tools and ...