Samuel
my favourite part is watching the frontend update in real time without having to touch anything
1 16
Samuel
ended up doing some fucked up ngrok things so you can use oauth in development. it's probably simpler than that but claude wrote a nice script that sets it up automatically
1 12
Samuel
next time I think I'm going to rewrite the backend to use the lexgen server? maybe? atprotomaxxing?
2 12
Why
Yeah, its been pretty solid. On friday i had it working on one project while i worked on another, periodically checking on it and unblocking it
1 16
austin
I had it write an entire project, soup to nuts, off some UML diagrams with pretty minimal guidance.
1 8
Why
Yeah, looking forward to them improving its abilities more too. Really wish it could google for docs and stuff online, or lookup deps i have imported
2 6
Samuel
it was grepping the types of atproto packages in node_modules to see how they worked, it was pretty cool
1 7
austin
yeah, itd be killer if it could work with a LSP. fyi you can use model context protocol stuff with it, so you can give it a tool that can go fetch docs and stuff
3
Billy J. Geiger
Hello, Austin!!!! "Newbie" here....I'm a head injury survivor (44+ years)....got hacked TWICE on Facebook, so I found this place....needless to say, I have NO experience in computer programming....how would I "join" a "group" in here?
1
Singularity's Bounty e/CC
IKR My company just sent out an email β€œToo many of you are using Claude. It is not approved. This must stop.” πŸ˜‚ The β€œthis must stop” was the clencher
4
Paul
Itβ€˜s super impressive, but tends to implement complex workarounds in cases where it doesn’t quite understand the syntax of a library. Even the concise setting doesn’t produce concise, easy to maintain code. Criticizing code quality rather than functionality speaks for the model thoπŸ˜…
3
Future RFK camper
Thanks for the recommendation !! So after reading that, am I correct to assume these 2 models are now outperforming GPT-4o and DeepSeek-R1 in general? I know it said Claude 3.7 is kicking ass for programming, but it didn’t mention how GROK 3 stands out. It only said it was trained w more power
1 3
Future RFK camper
I’ll definitely check out Claude !! Been wanting to learn more programming, but I’ve been waiting for AI to be good enough to where it can teach me instead of learning through more traditional means. For GROK I have concerns it might be biased, tho =\
1 2
πš’πš—πšπšŽπš›πš—πšŽπš πšπš›πš˜πš
I don't know how well it would be at the teaching portions, but I've never tried asking it for that. It's important to learn on your own so when it inevitably makes mistakes, you'll be able to notice & correct it.
GitHub - codecrafters-io/build-your-own-x: Master programming by recreating your favorite technologies from scratch.
Master programming by recreating your favorite technologies from scratch. - codecrafters-io/build-your-own-x
github.com
 1 3
Future RFK camper
Thanks for the resource! Currently I’ve been learning python, should I continue to stick to only one language? Would it be okay for me to explore? I’ve been told by a lot of people to just stick to one for a while, but I feel like I’d get a better understanding of how everything works by exploring
1 2
πš’πš—πšπšŽπš›πš—πšŽπš πšπš›πš˜πš
Relevant: "During the discussion, [Diana Hu, GM of YC] said that even if product builders rely heavily on AI, one skill they would have to be good at is reading the code and finding bugs."
A quarter of startups in YC's current cohort have codebases that are almost entirely AI-generated | TechCrunch
With the release of new AI models that are better at coding, developers are increasingly using AI to generate code. One of the newest examples is the
techcrunch.com
kittsycan.bsky.social
My only gripe with Claude is that it often overcomplicates and writes custom functions for functions that already exist in a package, which isn't great if you want to ensure standardized output.
Trabo
How does it compare with the builder mode in cursor or Trae?