I've been forcing ChatGPT into my life. It's been like learning a fun new word and trying to incorporate it into your vocabulary. You have to slow down a bit and overuse it. Sometimes it will not quite fit, but you learn something. And sometimes it serves you way better than the old clunky words, and you feel like you've grown.
Today, DHH shared a letter about How to Recover from Microservices in which he mentions Eric Evan's Domain-Driven Design and Martin Fowler's Patterns of Enterprise Application Architecture. I'm fairly familiar with both, but it's been a while. And I don't formally present those ideas like, ever. Why not do a refresher? Instead of going the traditional route -- google it, skim wikipedia and top blog posts -- I went directly to ChatGPT.
I am so glad I did. When I read anything, I jot down questions and notes and whatever little ideas pop up. Either in the margin or in my journal or a scrap of paper. But then what? If I'm not next to another CTO sorta guy, I've gotta dig deeper by doing more googling. Which I may not do. It's not super fun. With ChatGTP I was able to have a conversation!
It was (just) a machine-made pattern-matched dialog based on the existing cannon. Yet way better than reading (and dodging popups)! When you read to find an answer, the onus is on you to pull the answer from swamp. The author isn't there to help and the publisher just wants you to Subscribe Now. ChatGPT does the valuable work for you, increasing signal and dumping noise. It's fantastic. (PS: The experience of seeing the output as it's produced even feels better than reading a blog post that's been sitting there collecting dust.)
That being said, I am 1,000% sure the conversation would have been better with an experienced engineer. They would have volunteered important information I didn't exactly ask for, personal stories and maybe some juicy bits that have to be whispered. They would have asked me questions! Answering questions and sharing my own stories would have made the conversation deeper and stickier in my own mind. And the conversation would have deepened our connection as humans. But I'm in a coffee shop and the people next to me are amazing artists -- not engineers -- and they all seem pretty locked in to their work. 🤷♂️
For a little more nerdy fun, I also compared the GPT v3.5 answers to the v4 answers. The depth and breadth of the answers from v4 is impressive. So much better than 3.5 that I wouldn't even consider using 3.5 if I cared about learning something. Which I generally do. That being said, v4 is so much slower that it can get frustrating to wait for it. Not a real problem of course, since it'll be 10x faster by the time you read this (and they'll be on to v12).
Using ChatGPT to regurgitate old content for a "new" blog post is lame. How much of that do we need? Using it to reframe a concept and present it directly to me, to scratch the itch I have right now? Yes. This is a step change. I love it.
Today, DHH shared a letter about How to Recover from Microservices in which he mentions Eric Evan's Domain-Driven Design and Martin Fowler's Patterns of Enterprise Application Architecture. I'm fairly familiar with both, but it's been a while. And I don't formally present those ideas like, ever. Why not do a refresher? Instead of going the traditional route -- google it, skim wikipedia and top blog posts -- I went directly to ChatGPT.
I am so glad I did. When I read anything, I jot down questions and notes and whatever little ideas pop up. Either in the margin or in my journal or a scrap of paper. But then what? If I'm not next to another CTO sorta guy, I've gotta dig deeper by doing more googling. Which I may not do. It's not super fun. With ChatGTP I was able to have a conversation!
It was (just) a machine-made pattern-matched dialog based on the existing cannon. Yet way better than reading (and dodging popups)! When you read to find an answer, the onus is on you to pull the answer from swamp. The author isn't there to help and the publisher just wants you to Subscribe Now. ChatGPT does the valuable work for you, increasing signal and dumping noise. It's fantastic. (PS: The experience of seeing the output as it's produced even feels better than reading a blog post that's been sitting there collecting dust.)
That being said, I am 1,000% sure the conversation would have been better with an experienced engineer. They would have volunteered important information I didn't exactly ask for, personal stories and maybe some juicy bits that have to be whispered. They would have asked me questions! Answering questions and sharing my own stories would have made the conversation deeper and stickier in my own mind. And the conversation would have deepened our connection as humans. But I'm in a coffee shop and the people next to me are amazing artists -- not engineers -- and they all seem pretty locked in to their work. 🤷♂️
For a little more nerdy fun, I also compared the GPT v3.5 answers to the v4 answers. The depth and breadth of the answers from v4 is impressive. So much better than 3.5 that I wouldn't even consider using 3.5 if I cared about learning something. Which I generally do. That being said, v4 is so much slower that it can get frustrating to wait for it. Not a real problem of course, since it'll be 10x faster by the time you read this (and they'll be on to v12).
Using ChatGPT to regurgitate old content for a "new" blog post is lame. How much of that do we need? Using it to reframe a concept and present it directly to me, to scratch the itch I have right now? Yes. This is a step change. I love it.