Writing working snippets of code focused on a happy path scenario is easy (and doable by GPT-4).
Making a whole system and considering all kinds of edge cases is not.
One way we can tell that GPT-4 has not completely revolutionized programming today is how frequently chatGPT shows error messages during normal operation, asks that I refresh the page, etc.
It's a safe assumption its authors have considered using GPT-4 to improve their code.
https://t.co/SQcopjtb8H
“coding sucks”
When I hear something like this, I can't stop thinking that the author perhaps was never meant to be a software developer. If I think something sucks, I pay people who love their job to do it for me.
@annakaharris@carlorovelli@DavidDeutschOxf@Sara_Imari Perhaps too poetic to be useful, but I recall reading long ago the idea that each manifestation of reflective awareness is like a tiny shard of mirror the universe holds up to itself. With enough shards (via evolution), the universe can eventually see itself through itself.
While it’s great to have different ways to use GraphQL, once you discover (and understand!) Relay, there’s no way back.
What Relay really needs is better docs. And maybe better marketing?
Make Relay great again.
Every comparison of GraphQL vs whatever that fails to mention Relay is almost always useless. Even trpc comparisons completely fail to understand fragments. It's very easy to dismiss the complexity of GraphQL when you fail to understand the core value of the Query language.
How about we have “safe edition” for the sensitive readers, and keep the original as an “authentic edition”? Surely this would satisfy both sides. Or am I too naive, and sensitive readers expect everyone to be that sensitive?
I haven’t considered this perspective before (read the thread). I would have previously assumed building a yacht is better because it creates jobs and introduces new value into the world.
It is much worse for society if someone builds a yacht for hundreds of millions of dollars than if that person buys a vastly overpriced painting for the same price.
Imagine Michelangelo (or any other artist) being able to see revision history of their artwork, revert to any stage, branch out, collaborate with others and resolve conflicts.
Git commit messages add friction to creativity. Imagine Michelangelo sculpting the David, having to give a name & explanation to every dozen chisel hits.
Made an ESLint rule to ban certain forms of Date comparison (==, ===, !=, !==):
https://t.co/BAh10iyetK
Any ESLint plugin I can submit this to? A quick search didn't yield anything.
https://t.co/SCRKovgY91
I'd love to see this implemented on a CI system. Most of the overhead comes from having to perform set-up work all over again and writing data to/reading data from cache. This also increases the complexity drastically.
Imagine if we could split a running CI job into two. Every child job would maintain the perception that it's the original one. It's like Black Mirror, but for machines.