Greg.bot — A layman coding with Claude

Hi there,

For those of you who don’t know me, I’m Hamp Oldshue, and this is the blog I’m starting! I’m a recent graduate of the University of Alabama with an interest in AI. It’s clear that AI will revolutionize nearly every aspect of work, so I want to make sure I’m well-positioned to adapt to those changes. As a side hustle, I also buy and sell domain names. On this site, I’ll share my thoughts on AI, domain names, and life in general (as profound as that wisdom can be from a 23-year-old).

This past week, I’ve been using Claude to generate code for random projects I envision, just to gain familiarity and explore what it’s capable of. It turns out a whole lot! Now, I’m really not technical at all. I’m familiar with CSS and HTML for basic web design, but beyond that, I’ve never coded.

My experience with Claude quite frankly amazed me.

So I’ll start off with a rundown of what I’m trying to build and how things went. I’d seen several posts suggesting that an AI life coach or accountability partner could be successful, and I’d also read about the incredible user retention that Character.ai has achieved. If you could recreate the parasocial relationship that people (mostly teenagers, it seems) have with anime characters on Character.ai—but channel it in a positive direction to help people set goals and solve problems in their lives—I think it could be truly beneficial. That’s why I set out to build Greg.bot.

Not really knowing of Claude’s coding prowess I’d just started off prompting it on how I should go about coding this, but instead of giving me a roadmap it gave me a working python flask script that I then uploaded to VSCode. The original prompt was “How would I go about making an AI life coach that I could run on Claude’s API, that would off advice and allow me to tune it?” Then I prompted it to design me a minimal interface and I plugged in my Anthropic key and this is what resulted:

I later prompted it to create a more text-chat-like interface, aiming for a design that felt less like texting a chatbot and more like texting a friend—trying to eliminate the uncanny valley. I also asked for a responsive interface that worked seamlessly on both mobile and desktop, which Claude achieved in about two responses. Here’s the visual result. Much better:

Obviously there were some issues connecting the interface to the backend as I was getting back the error message I’d set. I’d realized that just uploading the files to my webhost and trying to deploy them there wouldn’t work. I was aware of both Github and Vercel but had never really used them before, so I read up on them and how to deploy the app using those. Being a layman I’d had some issues in setting the correct structure of the files, which I’m sure wouldn’t have happened had I been more technical. But I was able to get it resolved pretty quick and I was able to get a version of Greg.bot I liked.

This was all over the span of 2 days and hundreds of messages so it wasn’t as instantaneous as I made it seem, but it was still really quick and intuitive. Months less than it would’ve taken me to learn python and make an app of this complexity myself. The incredible potential of generative AI for code is that the non-technical layman, like myself, will be able build things they envision in a way they never could’ve without months of learning a coding language. And in a way after having interacted with the code, implemented it, and seen what it led to I have a much better understanding of Python and Java.

Anthropic I love yall but you have to get these fixed. I like to have long continuous chats and I need to know sooner that I’m about to hit my messaging limit so I’m not surprised by it. Very frustrating.

There were a few frustrations, like the messaging limits or the truncating of long code, but overall, it was a really great experience. Since Chat became popularized in late 2022, I’ve always thought AI’s best use case would be for coding. Anthropic has done the best job of addressing this that I’ve seen so far, and I’m incredibly bullish on their future. They’ve adopted an interesting strategy: identifying the highest-leverage use case for AI and going all in on that, instead of focusing on being generalists like OpenAI seems to have done with ChatGPT. They absolutely had to take this tact, as ChatGPT, as of right now, has a stranglehold on public mindshare. Anthropic has made Claude an almost essential tool, and I’m really excited for what comes next.

As for me I’m going to spend the next few days iterating on Greg.bot. I just added dark mode and light mode and I’m going to see if I can integrate Sms text somehow as well. Not sure whether this is a real go-to-market product or just a fun project to show people and see how far I can develop it, but if you’d like to check it out just go to the domain Greg.bot!

Please let me know what you think of this and share it if you like it or find it interesting! Shoot me a follow on Linkedin or Twitter. Or email me at Hamp@Hamp.com. I’d love to chat on anything from AI to College Football Playoff scenarios!