Hacker Newsnew | past | comments | ask | show | jobs | submit | quleap's commentslogin

Exactly. LLMs are a lot like human developers: they benefit from existing abstractions. Reinventing everything from scratch is a recipe for disaster—especially given an LLM’s limited context window.


I hate pip, a million times worse than conda


ADK is anything but minimal. It indeed has too much abstraction. It's certainly designed to promote Google's own services. If you are looking for lightweight frameworks, consider smolagents or Pydantic AI. Even OpenAI Agents SDK is simpler. LangChain/LangGraph are the epitome of incompetent designs.


I've also been on OneNote for 20 years. I've been trying out Obsidian. But there is simply too much stuff already in OneNote.


Your theory does not hold if a user initially had great experience for weeks and then had bad experience also for weeks.


ByteDance’s Volcengine is doing very well offering paid LLM services in China. Their Doubao Seed models are on par with other state-of-the-art models.


Claude Code had been supporting SSE transport without requiring mcp-remote.

So I don't really understand what's new in this announcement.

Maybe what's actually new is streamable HTTP and OAuth?

https://github.com/anthropics/claude-code/blob/main/CHANGELO...

  1.0.27
  Streamable HTTP MCP servers are now supported
  Remote MCP servers (SSE and HTTP) now support OAuth
  MCP resources can now be @-mentioned


I'm extremely satisfied with Pixi. It fixes almost all the issues I had with conda and mamba. It supports both conda and pypi (via uv) packages. I don't know if uv fixes pip's dependency management hell. I settled on conda packages because pip was such a mess.


not true in my experiments


Mercurial, with its changeset evolution and phase management, is gazillions times more powerful than Git in terms of history rewriting.


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: