I mean, how many times have you told AI not to do something, then a minute later seen it do that exact thing again? If I had to guess, a lot. But here's what I've figured out. When you write "don't" or "do not," the AI can easily read that as "do," which means it skips right past the "don't." This is why I avoid negatives and use positives like this: - "Output plain text only" - "Respond in 2 sentences max" - "Limit answers to features and support topics" - "Answer only from the provided context. If it's not there, say 'I don't have that information.'" - "Respond in Japanese" - "Begin your response with the first data field" These avoid the main issue of the AI misunderstanding you, while making sure it actually does what it's supposed to. Try this next time you're about to write a hard rule. You might get surprised.

I jumped on a call a few days ago to help someone in the community with n8n. He shared a problem, a service list that wasn't accessible online. The solution he found was to take screenshots of all the items and share them with ChatGPT. But I knew of a simpler solution. Scrape the website, turn the results into structured JSON data, and use an API to get it. Much simpler, and zero manual work. But this was only the start. I then fired up a server on Digital Ocean and made the API public with an API key, so only he could use it for his clients. Here's the thing, this is super useful for him because now he has live data on all services with no manual work to get it. As I'm writing this I'm building a YouTube comment scraper for an Ecom brand. It will run live on the server and scrape comments every 6 days. Super useful since we don't need to go looking for comments. We scrape all comments, save the link, and comment directly under them.

Tell me if you've seen this before. "We need an n8n automation expert. You must have hands-on experience with [random tool], REST APIs, and webhooks. 10+ years with AI required." AI is five years old. Nobody has ten years with it. So what's your first thought? "I can't do this. I've never touched that tool. I don't have ten years of anything." So you skip the post and move on. You just did yourself a disservice. Because if you hopped on a call with that business owner, you'd find out fast: the thing they want solved is way easier than the post made it sound. I see this every day. A hire post stuffed with big words, asking for the world. But the real work behind it is simple. I used to work in a pure code base with a basic understanding of what I was doing. That was enough to ship working code. And with AI now? You can do this. So here's what I'm saying. Stop rejecting yourself off a job post. Get on the call. Talk to them. Worst case, you learn it's too hard and you pass. Best case, you realize you can do it. Most of the time, you can.

Has anyone ever used firecrawl to scrape Reddit? For example do routine passes looking for specific keywords or phrases so I can go reply.

So I've seen a few posts about people complaining about Claude. I'm not sure what's going on. But Claude is working crazy good for me, and the reason why might be because of your setup. So to help out, I'll share how I approach working with these AI models when coding. - Claude should always make the least amount of changes, no matter what. Don't let Claude make 10+ changes in one run. - Run tests on the changes that have been added, end-to-end. Ask Claude to test everything so it works. - Ask Claude, what could go wrong here? What are we missing after critical changes? - Run sub-agents for tasks and keep your main context window as small as possible. - Make sure your claude.md file isn't bloated. A bloated one will cause so many problems. Try these things out and let me know how it goes!