AI will not replace you. A person using AI will.

Software development has never been about writing code. This was obvious to many before, but now AI has made it crystal clear to everyone. Building software is 99% thinking, 1% writing code, and 1% fighting off-by-1 errors.

Web scraping + Agents in a single FREE session! This is a FREE, live session for anyone interested in learning about web scraping, agents, and automations. This is happening NEXT week: Thursday, October 23 10 a.m. PT / 1 p.m. ET Some of the topics: • How to use Apify's MCP

I recorded my 4 agents working. It's a multi-agent game with 4 different agents interacting with one another: • Tournament agent • Game agent • Player agent • Commentator agent You can see the back and forth from both players while they compete. I built this using Google

Web scraping + Agents in a single FREE session! This is a FREE, live session for anyone interested in learning about web scraping, agents, and automations. This is happening NEXT week: Thursday, October 23 10 a.m. PT / 1 p.m. ET Some of the topics: • How to use Apify's MCP

AI pilots prove intelligence, but AI in production demands reliability. The best teams separate their stack early: 🧠 Core = how AI thinks 🛡️ Reliability = how it stays safe That’s how prototypes become products. 👉 https://t.co/JtOO6rpKhV

I recorded my 4 agents working. It's a multi-agent game with 4 different agents interacting with one another: • Tournament agent • Game agent • Player agent • Commentator agent You can see the back and forth from both players while they compete. I built this using Google

Today, the cost of trying new things is zero (or near zero). This is one of the best outcomes of AI coding. Before, if you had three different ideas to solve a problem, you had to pick carefully. Most people usually picked the "safest" idea because the cost of trying it out and

Tools don't write good software. You do.

You can build websites with relatively little training. It’s much harder to build machine learning models. I built my first website after two weeks of learning+coding back in 2000-ish. It wasn’t a masterpiece, but it was functional and I got paid for it. It took me several

Still, one of the most frustrating experiences with AI coding is asking for one change and getting back 10 random updates you didn't ask for. This happens to me every day, and it's frustrating beyond belief. Every model does this, but I find GPT-5 is the worst offender. The

More than 3,000 students have taken my AI/ML Engineering program! This has been life-changing to many of them. Read some of the latest feedback I received from the last cohort. The next one starts in 3 weeks, on November 3. This class will be different from anything else

You don't need Claude Code. You don't need Cursor. You don't need Codex. All of these tools are great, and there's a ton of overlap between them, so you need to pick one and build some mastery. People who constantly jump from one tool to another are taking advantage of only

With the model, you can: • Combine up to 6 images • Match their style, objects, characters, and poses • Generate readable text • Update specific regions of an image • Generate 4K output results Here is the link to check out the model: https://t.co/Ti4yRJC8lv Thanks to

China is beating the US on AI. ByteDance's Seedream 4.0 is currently the #1 text-to-image generation model, surpassing Google's Nano Banana. It's really good! The model is available on ByteDance's platform, Dreamina. I recorded a quick video. See it in the next post.

Many people believe *most developers* use Claude Code or similar to write their code. This place is a bubble. Go outside and most developers have never even heard about Claude Code and are using ChatGPT to write dinner recipes.

I’ve noticed this as well: Always ask for a plan before letting the agent work on your code. It’s tempting to let agents go gung ho right away because you think you’ll save some time, but results will almost always be worse. 1. Ask it to generate a plan 2. Review and correct

Claude Code and Codex do not replace Copilot and Cursor. I've already heard multiple people make this argument, and I think it comes from the vibe-coding community because of the way they use these tools. First, Claude Code and Codex are agentic coding tools. They are good at

If you buy an expensive camera, your photos will suck as much as before. If you buy an expensive golf club, you won't start swinging 300-yard bombs overnight. It's not the tools. It's the person using them. Building good software has nothing to do with whether you use Claude

Tools don't write good software. You do.

More than 3,000 students have taken my AI/ML Engineering program! This has been life-changing to many of them. Read some of the latest feedback I received from the last cohort. The next one starts in 3 weeks, on November 3. This class will be different from anything else