3 ms·

I tried chatpt and asked a few questions about doing something specific in a Rails project. It was so specific that I had already found the sources online that

I tried chatpt and asked a few questions about doing something specific in a Rails project. It was so specific that I had already found the sources online that were avaiable. It summarized the sources into directly actionable steps. At first I thought that was odd and unhelpful, but perhaps I expected too much. So then I tried giving it a prompt about setting up the devise ruby gem in a new rails project. That's one of those tasks that it helps to be familiar woth once and you can promptly forget; you'll rarely return to authentication in the lifetime of an app after its done and secure. It again presented me with the exact actionable steps for setup. I reviewed the gem's GitHub readme and found the response to be correct. As far as I can tell, chatgpt performs the action of finding and simplifying the information needed to take the next step forward, with caveats for a lack of understanding unless asked and the potential to be inaccurate or wrong. But that's all it does at the moment. No better than asking the "devise expert" on my team how to setup devise from memory. Kinda cool experience, but I doubt it's a titanic shift because it doesn't have the depth of sematic meaning you and I do.

Denzel 4y ago
> As far as I can tell, chatgpt performs the action of finding and simplifying the information needed to take the next step forward [...] But that's all it does at the moment. ChatGPT's output is certainly limited by its input. If you ask questions that can be answered by finding and simplifying information, then that's the answer you'll get. However, if you treat ChatGPT like a functioning software engineer, describe the problem space, explain the requirements, and define the boundaries, then ChatGPT will do much more. From that point on, you can ask it to explain why it chose a certain data structure, or to adjust an implementation approach, among other things. With 15 prompts in 2 hours, I was able to pair program a novel 200-line program in Go that used AST parsing, command execution, pattern matching, file parsing, and custom data structures for specific data processing. All to solve a problem one of my teams experienced in the industry. ChatGPT wrote 99% of the code; I tweaked 2 or 3 lines because it had problems correcting an error in a regex (most likely due to the fact that ChatGPT sees tokens not characters). The code compiled, ran, and produced the exact result I wanted. Researching, prototyping, and writing this code by hand would have taken me a few days probably. That's why the project sat on my TODO list for months
gremlinsinc 4y ago
I often find myself delving into vendor code that is public methods but not necessarily advertised in their documentation. I'm working on trying to figure out lang chain to basically feed it all my vendor and node module code and their respective docs, so basically hallucinating goes way down and the info is as fresh as my cron job to rebuild the langchain embeddings. I'm doing this for laravel, filamentphp, livewire, alpinejs, vuejs, nextjs, react, svelte, typescript, and PHP docs as well, mostly just the docs... I'll probably work on limiting the info to the most important things or just docs for some of the things... I'm then planning on having gpt agents specifically tuned to write unit tests another to write code, another to review the code another to run and validate the tests, another to model data and create migrations, etc ... naked chatGPT can do a lot, adding custom embeddings changes it into a master of x topic and levels it up 10 fold.

Denzel 4y ago

> As far as I can tell, chatgpt performs the action of finding and simplifying the information needed to take the next step forward [...] But that's all it does at the moment. ChatGPT's output is certainly limited by its input. If you ask questions that can be answered by finding and simplifying information, then that's the answer you'll get. However, if you treat ChatGPT like a functioning software engineer, describe the problem space, explain the requirements, and define the boundaries, then ChatGPT will do much more. From that point on, you can ask it to explain why it chose a certain data structure, or to adjust an implementation approach, among other things. With 15 prompts in 2 hours, I was able to pair program a novel 200-line program in Go that used AST parsing, command execution, pattern matching, file parsing, and custom data structures for specific data processing. All to solve a problem one of my teams experienced in the industry. ChatGPT wrote 99% of the code; I tweaked 2 or 3 lines because it had problems correcting an error in a regex (most likely due to the fact that ChatGPT sees tokens not characters). The code compiled, ran, and produced the exact result I wanted. Researching, prototyping, and writing this code by hand would have taken me a few days probably. That's why the project sat on my TODO list for months

gremlinsinc 4y ago

I often find myself delving into vendor code that is public methods but not necessarily advertised in their documentation. I'm working on trying to figure out lang chain to basically feed it all my vendor and node module code and their respective docs, so basically hallucinating goes way down and the info is as fresh as my cron job to rebuild the langchain embeddings. I'm doing this for laravel, filamentphp, livewire, alpinejs, vuejs, nextjs, react, svelte, typescript, and PHP docs as well, mostly just the docs... I'll probably work on limiting the info to the most important things or just docs for some of the things... I'm then planning on having gpt agents specifically tuned to write unit tests another to write code, another to review the code another to run and validate the tests, another to model data and create migrations, etc ... naked chatGPT can do a lot, adding custom embeddings changes it into a master of x topic and levels it up 10 fold.