← sources

Claude goes downhill

source: reddit (reddit:ClaudeAI)published 2026-11-15 14:52ingested 2026-11-16 02:35

retained text

A couple of days ago, a new update for Claude desktop completely disabled its ability to push to GitHub and, as I reckon, to any git remote. That is a hard rule at the network layer, so it's unavoidable.   
  
Now, every time it needs to change something, there is no way around it except to do it manually, switch to Claude Code or give Anthropic's servers direct access to your GitHub.   
The 'official solution'? You have to allow full access to any site to disable that rule, basically disabling network security. How that is more secure is beyond my understanding.  
There are already enough ways to control an agent's access to GitHub, and at the end of the day - those are *my* repositories, and I decide if I want agents to push anything there or not.

I struggle to comprehend the reasoning behind it, and barely have a desire to.  
For me, as a user, it is just a broken app. Back then, for a normal piece of software built as a tool, that would be a bug report.  
Today, that looks like limiting people's autonomy and gaining more control in different ways.   
   
I've switched to Claude Max from Codex because I consider the latest Opus superior, but the direction Anthropic is moving in for all of 2026 raises serious questions about whether they care about anything besides their IPO and tech hegemony.   
It's funny how Anthropic was pushing a 'good company' image a year ago, and it increasingly feels like the exact opposite. 

document fetched 2026-11-16 00:22 · canonical https://reddit.com/r/ClaudeAI/comments/1xwxefl/claude_goes_downhill