i dont think youd need 671b for something like this
this seems to be a command detection which 8b should be able to do just fine
45 karma · joined March 7, 2017
this seems to be a command detection which 8b should be able to do just fine
it should be a simple config change tbh. would love to merge a pr for that. else, will try borrowing a mac & iphone over the weekend :)
Many times, i make commitments to people - like, i will get back or i will follow up. But then that just gets lost in all the noise. I forget where i made those commitments.
So, i built this simple tool where each time i make such a commitment, i can just take a screenshot and share it with this app. the app will create a todo which i can mark as done later.
super simple app. do give it a swing :)
the other thing i wanted to share is that i thought this was a trivial enough app for chat gpt to make. And that's why i signed up to build this at 1am. i thought i'll be in and out quickly. I got chat gpt to write the entire code.
But ...
It didn't work. I spent hours debugging. and the errors were so small that i didn't spot it at all!
particularly this one : I needed other apps to be able to share an image with this app. On the android manifest, i needed to set a configuration called "android.intent.action.SEND"
but chat gpt gave me the code which had "android.intent.android.intent.action.SEND"
I didn't spot it at all and went down several rabbitholes to figure out why the configuration isn't being set. I thought it was a compiler issue so i did several rounds of clean builds and decompiled the build to check the configurations.
It was a nightmare.
I gave up and went to sleep.
Today morning, i rewrote the app from scratch. It took me 30 minutes. I still used chat gpt, but only for very specific mundane/boilerplate code. Everything else i wrote myself.
I find these AI tools to be incredibly useful to code, but i they are not able to build e2e systems. Even if they build something that's 80% there and i have to write only the 20% that'd be great. but i fear that's not where the state of the art is. It significantly sets you behind if you get it to write large chunks of code.
If digital artists can have their own NFT Marketplace to sell digital paintings NFTs, hackers deserve the same.
This is a fun side project we built. We don't make money from this. We just want to make sure there is a neat way to reward the open source contributors.
Though this is a sideproject, I'm keen on making this successful. It's currently launched on a testnet. I wanted to have comments from the HN network to see how we can make this better before we launch more generally.
What do you look for in a candidate at the screening stage?