3 ms·

It's not about building a product, it's about saving time by making it do small task 30x faster than yourself (just like author did). Not all tasks, but many ma

It's not about building a product, it's about saving time by making it do small task 30x faster than yourself (just like author did). Not all tasks, but many many menial tasks. At least that's how I use it and I'm super happy with what it gives me.

taude 3y ago
Agree with this, not sure why the parent of this thread thinks you'd use it to output a full business. But helping write at the functional level is game changing: I've built entire GitHub Actions to help improve team effeciencies. Built validation rules. A bunch of Python scripts. Parsed a multi-level JSON response that would have taken a few hours, etc... All in minutes. Imagine knocking off like 15 of these tasks before noon.
- test6554 3y ago
  I am worried that Chat-GPT productivity levels will become the new norm and when it doesn’t give you what you need you will be underperforming at work.
dmje 3y ago
That's exactly right. Any product is essentially made up of a gazillion small tasks, whether setting up a framework or writing a database query. If just some of those small tasks can be made quicker / easier using these tools, then why not? What I find most useful in this space is twofold: 1. Specificity: so yeh, I can look at SO or wherever and see responses to a semi-specific query, say about querying a database. And yes, I could post a question and maybe after a few hours if I'm lucky get a response. But with ChatGPT I can throw in the actual table and column names and get a specific response back. Most often it needs honing, sure, but it's most often a great starting point. 2. Being able to back-query. So the fact that history is maintained means I can say "actually, add in a column called X and can you make sure that the query only returns unique IDs" and - again - I'll get an immediate response without having to go all the way back through the questioning journey again. As with any tool, the extreme responses on either end seem obviously flawed. On the one hand: no, it's not going to replace all developer jobs. On the other, to just completely ignore it, not try it, always assume that "the documentation" is better - this seems blinkered and foolish too. Somewhere in the middle is a sweet spot where human brains collide with these tools and we manage to tease out code that is more quickly researched and saves time but also well considered and treated with a degree of healthy scepticism before deployment.

taude 3y ago

Agree with this, not sure why the parent of this thread thinks you'd use it to output a full business. But helping write at the functional level is game changing: I've built entire GitHub Actions to help improve team effeciencies. Built validation rules. A bunch of Python scripts. Parsed a multi-level JSON response that would have taken a few hours, etc... All in minutes. Imagine knocking off like 15 of these tasks before noon.

test6554 3y ago
I am worried that Chat-GPT productivity levels will become the new norm and when it doesn’t give you what you need you will be underperforming at work.

dmje 3y ago

That's exactly right. Any product is essentially made up of a gazillion small tasks, whether setting up a framework or writing a database query. If just some of those small tasks can be made quicker / easier using these tools, then why not? What I find most useful in this space is twofold: 1. Specificity: so yeh, I can look at SO or wherever and see responses to a semi-specific query, say about querying a database. And yes, I could post a question and maybe after a few hours if I'm lucky get a response. But with ChatGPT I can throw in the actual table and column names and get a specific response back. Most often it needs honing, sure, but it's most often a great starting point. 2. Being able to back-query. So the fact that history is maintained means I can say "actually, add in a column called X and can you make sure that the query only returns unique IDs" and - again - I'll get an immediate response without having to go all the way back through the questioning journey again. As with any tool, the extreme responses on either end seem obviously flawed. On the one hand: no, it's not going to replace all developer jobs. On the other, to just completely ignore it, not try it, always assume that "the documentation" is better - this seems blinkered and foolish too. Somewhere in the middle is a sweet spot where human brains collide with these tools and we manage to tease out code that is more quickly researched and saves time but also well considered and treated with a degree of healthy scepticism before deployment.