|
Hey Reader, Opus 4.7 dropped last week. GPT-5.4 is probably getting an update this/next week. So let’s talk about what I actually pay for, what I actually use, and what you should do if you can only pick one. Quick answer: the best setup is still both. Claude AND Codex. But that’s not realistic for everyone, so let me break it down. Claude (Opus 4.7).Best all-rounder. Jack of all trades, master of none. Code? Pretty good. If you can only afford one subscription, this is it. But skip the $20 Pro plan. Go minimum Max 5x or don’t bother. The $20 tier will leave you rate-limited and frustrated by from day 1. Heads up: no image generation. And they ship so fast their QA is basically you. Bugs happen weekly. Still, Claude.ai and Claude Code are both genuinely excellent. For most people reading this, this is your pick. Codex (GPT-5.4).Codex is the best coder in the world. Full stop. But the ChatGPT product around it is kind of trash. Every time I open it I get annoyed. It hallucinates and is borderline useless. Exception: GPT-5.4-pro is the smartest research model on the planet. Nothing touches it for deep thinking. But Codex cannot design to save its life. Your apps will look like a 2004 Craigslist page. And I’d never write an email like this one with it. So for me, Codex is the complimentary subscription. I write most of my production code with it because it’s more reliable and more trustworthy than anything else. Most of my clients do the same. Think of Codex as the quiet engineer in the back who ships clean work but can’t pick out a shirt. Claude is the designer-writer-generalist up front. Cursor.This one confuses people because of the price. Cursor is expensive. Like $300, $400, $500, sometimes $1,000+ a month. Especially if you’re running the best models without watching the meter. But here’s the thing. Cursor has the best user experience in AI coding, period. You don’t switch between tools. Cursor orchestrates the models for you. Best Claude integration and smoothest MCP/Skills plugin system. A massive community with real events in real cities every week. It’s like hiring a general contractor instead of managing every sub yourself. It does cost more (a lot more), but i does save your sanity. Not worth it for solo hobbyists. 100% worth it for businesses, teams, or anyone whose time is worth more than their AI bill. What YOU should use.
That’s my April 2026 recommendation. Copy it, tweak it, ship something. Talk next week, P.S. Opus 4.7 is a few days old now and I have a lot of thoughts. I’ll stress test some more, report back in my next email. It supposedly is a lot better with creative things like designing and writing. It wrote this email. Did you notice? |
Coder of 20+ years teaching non-technical people how to build their own software business in 30 days with AI. No devs or code required.
Hey Reader, I have a confession. For the last few weeks, I barely worked on my business. Here's what happened. When the Middle East situation escalated, I left Dubai. Since then, I've been all over the place. Different countries, different time zones, zero routine. I kept delivering to my AI Architect Program clients, but everything else was in maintenance mode. And almost the entire rest of my day? I spent it on Twitter. Watching the AI wars. Reading every hot take. Getting genuinely angry...
Hey Reader, Imagine paying thousands of dollars to an "AI expert" to build something for you. And then they deliver... absolutely nothing. That is exactly what happened to a good friend of mine. He hired a "reputable" AI expert to automate his agency's design workflow. They were promised automatic design-to-code conversions, a full system, the works. This would cost thousands of dollars and take a few weeks. I really wish he had talked to me before signing. But he didn't. He only called me a...
Hey Reader, Nobody wants to say this out loud, so I will. The "free AI tools are good enough" era is ending. And if you've been waiting for the perfect cheap tool before you start building your idea... the window is closing faster than you think. What I Found Last Week I tested Kimi K2.5, one of the most hyped "free" AI models right now. Won't lie to you, it's actually decent. The closest to Claude of all Chinese models. But here's the thing nobody is talking about: The gap between free AI...