SharedSourceAI's profile picture. Your compass for the ever-evolving world of Open Source AI. We explore cutting-edge GitHub repos, demystify complex models, and spark discussions on the future

Open Source AI

@SharedSourceAI

Your compass for the ever-evolving world of Open Source AI. We explore cutting-edge GitHub repos, demystify complex models, and spark discussions on the future

고정된 트윗

Turns out Devin skipped the Zoom calls, so it never got promoted.


Every startup is racing to launch an ‘agent framework’. But the real power isn’t in frameworks—it’s in complex agents that can actually do things. Coding agents like Claude Code, Codex, Factory proved it. The next wave? Bringing that power outside coding. Huge whitespace.


Open Source AI 님이 재게시함
trq212's tweet image.

Why is he bullshitting so much these days?

I'm building an army of tools to automate my life. That's the good part about vibe-coding: you can automate anything you want in a fraction of the time it would have taken a few years ago. There is an ocean of vibe-coding tools, but here is one that does something different:…



The power of token prediction

nothing... LLMs are just a token prediction

slow_developer's tweet image. nothing... LLMs are just a token prediction


Chaos hires better

Companies should recruit on X, not LinkedIn.



Copilot giving us OpenAI + Claude in one package is 🔥 … only thing missing now is an SDK, especially after getting a CLI. Imagine the custom dev possibilities 👀🚀 #AI


Claude was flexing on “agentic coding,” Codex brought CLI magic, but Copilot’s like: “I got editor integration, baby!#AI

SharedSourceAI's tweet image. Claude was flexing on “agentic coding,” Codex brought CLI magic, but Copilot’s like: “I got editor integration, baby!#AI

Your MVP database journey: Day 1: 'We need MongoDB for scalability!' Day 30: frantically googling 'how to do JOIN in MongoDB' Day 60: quietly switches back to PostgreSQL


Open Source AI 님이 재게시함

🔥 The only agentic coding pattern you need. Add one thing: 4. Verification - Create feedback loops to test that the plan was implemented correctly (models still occasionally fail on execution) Bookmark this and copy into AGENTS .md From @ericzakariasson

daniel_mac8's tweet image. 🔥 The only agentic coding pattern you need.

Add one thing:

4. Verification - Create feedback loops to test that the plan was implemented correctly (models still occasionally fail on execution)

Bookmark this and copy into AGENTS .md

From @ericzakariasson

You won’t believe what people are saying now: Gemini 2.5 Pro & Sonnet/Opus 4 are good… but GPT-5 just destroyed them. Wait till Google & Anthropic try to catch up.


When your LLM agent goes multilingual mid-answer with random Thai + Chinese chars… that’s not “creativity,” that’s a TPU bug

We've published a detailed postmortem on three infrastructure bugs that affected Claude between August and early September. In the post, we explain what happened, why it took time to fix, and what we’re changing:



United States 트렌드

Loading...

Something went wrong.


Something went wrong.