Mohamed Alrujayi

88 posts

Mohamed Alrujayi banner
Mohamed Alrujayi

Mohamed Alrujayi

@aburjg

@CyclsAI | ...

Katılım Şubat 2024
95 Takip Edilen19 Takipçiler
Mohamed Alrujayi retweetledi
rahul
rahul@rahulgs·
launching nanocode! minimal claude code implementation. zero deps, ~250 lines of python. full agentic loop with tools (read, write, edit, glob, grep, bash). prompt is just "concise coding assistant. cwd: /path"
rahul tweet media
English
67
57
1.1K
194.8K
Brian Cardarella
Brian Cardarella@bcardarella·
Claude Code adding itself to every fucking commit as a co-contributor is pretty obnoxious
English
210
49
5.8K
439.6K
Mohamed Alrujayi
Mohamed Alrujayi@aburjg·
Ralph Wiggum mode is the first real glimmer of hope for autonomous agents. $200 in API costs → $50k contracts delivered. Perhaps we've been overcomplicating this. The answer wasn't a better planner or a smarter model. It was a dumb while loop burning tokens.
English
0
0
0
54
Mohamed Alrujayi
Mohamed Alrujayi@aburjg·
Markdown is the hottest programming language rn
English
0
0
0
107
Mohamed Alrujayi retweetledi
jacob ۞
jacob ۞@jehovahscript·
Coding with Claude
GIF
English
137
873
10.1K
941.6K
Mohamed Alrujayi retweetledi
Mohammed محمد
Mohammed محمد@Mo_tech_·
I finished a travel app with the help of claude 3.5 in 2 days :) full code on @Replit: @mf80/RAG-and-function-calling-example?v=1" target="_blank" rel="nofollow noopener">replit.com/@mf80/RAG-and-… Tool used: - @AnthropicAI claude 3.5 for coding - @cycls_ for streaming the web app - - @Replit for running the app - @OpenAI model+API - @trychroma vector db
English
0
1
9
831
Mohamed Alrujayi retweetledi
Amjad Masad
Amjad Masad@amasad·
The real test of an AI Software Agent is whether it can create software faster than it would take the user to google an app for the exact use case.
English
30
18
337
32.3K
Mohamed Alrujayi
Mohamed Alrujayi@aburjg·
@amasad Absolutely, AI coding assistants are really good at writing code, not creating full apps
English
0
0
0
60
Amjad Masad
Amjad Masad@amasad·
Claude is surprisingly good at one-shotting frontend apps but it's typically hard to figure out how to setup the environment and deploy the apps -- Replit fixes that.
matt palmer@mattyp

Last week, we introduced support for @AnthropicAI's Claude 3.5 Sonnet in @Replit. What does that mean? You can create and deploy Artifacts on Replit without writing any code. Here's a walkthrough of a sample app I built and deployed, without leaving Replit, with only basic framework knowledge and prompting. If your curious, it uses @reactjs, @shadcn, @vite_js and @tailwindcss AI-native development with deployment frameworks (like Replit) is the future for people who want to ship.

English
16
22
374
41.3K
Mohamed Alrujayi retweetledi
Dennis
Dennis@dkardonsky_·
@sama Mistral and meta: ship models OpenAI: ship blog posts
English
7
10
319
10.9K
Cycls
Cycls@CyclsAI·
Hello world app in 6 lines of code, as it should be
Cycls tweet media
English
1
1
4
647
Mohamed Alrujayi
Mohamed Alrujayi@aburjg·
The right question for AI yields magic
English
0
0
0
179
Mohamed Alrujayi
Mohamed Alrujayi@aburjg·
@amasad Or perhaps that the conversational interface is more natural to the brain than the IDE
English
0
0
1
67
Amjad Masad
Amjad Masad@amasad·
What’s fascinating about this trend is that many users like this one prefer to fully generate/iterate on the code in Claude App as opposed to IDE extensions—even those with builtin Sonnet support—then copy-pasting to the editor. Suggests that LLM quality trumps packaging and UX.
Garry Tan@garrytan

Underrated trend

English
48
23
616
85.2K