Content pfp
Content
@
https://opensea.io/collection/dev-21
0 reply
0 recast
2 reactions
Rafi pfp
Rafi
@rafi
Frameworks make much less sense now since much boilerplate can be generated in seconds
1 reply
1 recast
5 reactions
Phil Cockfield pfp
Phil Cockfield
@pjc
Strongly agree. At the library level too. Being able to shed significant upstream dependencies, on well understood and widely implemented design patterns, by inlining rapidly generated, and solidly tested library code, is gonna be more and more common in my future i reckon. I already did it a few times on things, and it gave me pause how well it worked.
1 reply
0 recast
2 reactions
Mr. Wildenfree 🐺🍡🎡 pfp
Mr. Wildenfree 🐺🍡🎡
@mrwildenfree
Do you mean this in regards to like AI generated code? And this working more effectively than something like Vue.js or something of the sort?
1 reply
0 recast
0 reaction
Rafi pfp
Rafi
@rafi
Yes AI. Vue maybe not cause it's too general. I had more specialized and narrow libs in mind, like data processing or graphics
2 replies
0 recast
2 reactions
Phil Cockfield pfp
Phil Cockfield
@pjc
Yes, I was referring to AI assisted generation. I'm cautious on expressing any opinions too definitively because the landscape is change, and fast. I'm personally currently of the approach of talking with the LLM, generating code paths, re-writing then, hand unit-testing them so I can prove to myself I actually know what just happened, and I'm not introducing subtle bugs etc etc. Kind of the opposite of what the brand new term "Vibe coding" is pointing to. To make my comments a little less abstract, the epiphany I had and was referring to above was when I had some stupid ESM import problem with a really really old library for parsing cmd-line args (`minimist`) .. and Deno's std lib for that was also being a dick, and failing on imports. Generating that kind of thing up from nothing, while always possible, had a much higher cost, especially in the area of "confidence" that you actually hit the main edge cases the library was abstracting for you.
2 replies
0 recast
1 reaction
Phil Cockfield pfp
Phil Cockfield
@pjc
All this "old code" is part of the training set now - so new possibilities are emerging here. "Find the dependencies and illuminate them" - the battle cry of the early Excel team* *ref: https://www.joelonsoftware.com/2001/10/14/in-defense-of-not-invented-here-syndrome/
0 reply
0 recast
1 reaction
Mr. Wildenfree 🐺🍡🎡 pfp
Mr. Wildenfree 🐺🍡🎡
@mrwildenfree
Fascinating approach, and this sounds like an ideal way to approach it, learning & reverse engineering it along the way of the build process. I have limited use cases at the time, and I know that technically the AI is the worse it’ll ever be right now, so I always wrestle with what is important for me right now as a lyricist & music artist first. I know I could get pulled into a heavy gravity of deep dives in computer science, coding/programming, etc., yet I’m wanting to make sure I’m not diluting myself or my current efforts in this season. There’s a specific escape velocity I’ll need to hit with the music first!
1 reply
0 recast
1 reaction