Title says it all. I watched an introductory video on using Cursor, Claude, and Replit to build a website, set everything up, and after a lot of trial and error, I now have a fully functioning software product.
It’s an email verification tool that cleans and verifies email lists. Nothing groundbreaking, but definitely one of the cooler things I’ve built so far, and I’m hoping to get some customers.
I built it with React and Node.js technologies I hadn’t heard of before starting.
Some notes from the experience:
- While AI handled most of the coding, it took over 50 hours of back-and-forth interactions to get things right, fix mistakes, and unbreak the code after some misunderstood prompts.
- Sometimes I felt like my codebase was overwhelming Claude with too much context, so I switched to using Anthropic’s website for a contextless chat. This often solved the problem when Claude was repeatedly getting things wrong.
- AI struggled with CORS issues. As someone unfamiliar with CORS, this was incredibly frustrating to troubleshoot.
- AI also wasn’t great with external APIs due to outdated documentation. I eventually learned to feed it the correct docs using Cursor or manually pasted the latest docs. Even then, it would sometimes revert to its older knowledge.
- Occasionally, Claude couldn’t solve a problem, but GPT could—and vice versa.
- I wish it could integrate tools like MidJourney or other native image generation tools within Cursor to create visuals for my website and blog. I haven’t had the chance to tackle that yet.
As part of the project, I built a script that reads RSS feeds from my Medium articles, repurposes and rewrites them, and then posts them to the blog once a day. My assistant handles the Medium content, so we’re leveraging the RSS feed. (This could be an app on its own, actually.)
I’m currently working on a Chrome extension and a Google Sheets plugin with the same functionality. More updates coming soon!
I am familiar with the complexities of email verification systems because I have constructed my own. I’m hoping it’s okay if I ask a few questions since it took me a while to figure out while I was creating it. If it had been available to me a few years ago, it would have helped me grasp the current status of AI technologies.
Is it capable of handling every domain, such as yahoo, outlook, Google.com, and so on, without being referred to as catch-all?
Is greylisting handled by it?
What steps do you take to keep yourself off the blacklist?
Can you confirm emails from domains that frequently return 429, such as “Too many requests”? I’ve observed that, when done right, these domains can be verified.
I’m not bothered at all!
Because I can’t figure out how to read the email, I can’t seem to get Yahoo and AOL to operate. The email can exist, but the inbox might be closed. Unfortunately, it will only yield catch all if a domain is a catch-all. The only remedy I could find was to send real test emails.
Indeed, it does.
I’m still not quite there. That’s about it; we’re using cached results, we’ve implemented rate restriction and throttling, and I’m setting up proxies.
After a 30-minute trial, we’ll mark the request as unknown. I might also attempt proxies in this case.
I hope that’s useful.
Many thanks. The code is superior to most codebases because it aids in greylisting and retries. Best wishes to you!
USB with relation to point 1. Did you manage to find a way to confirm emails that belong to a catch-all domain?
I can use Google, Outlook, Yahoo, and so on, but not all of them. In general, the majority of personal email addresses. I took advantage of a pattern that these all had in common. Since it is constantly changing for the remaining domains, I was unable to discover a means to confirm them.
It appears that you have boarded a rapidly moving train, my friend. That’s also what I’m attempting to do. My only problem is that I’m still researching the best kind of micro SaaS to create. There are countless possibilities. As such, it can be a little overwhelming to even start.
In these situations, no-code platforms considerably facilitate the productivity gains of non-tech founders by providing basic SaaS apps fast.