I'm Passing on theGitHub CoPilot CLI Challenge

Collapse
X
 
  • Time
  • Show
Clear All
new posts
  • MyrinNew
    Senior Member
    • Feb 2024
    • 5175

    #1

    I'm Passing on theGitHub CoPilot CLI Challenge

    I'm an advocate of AI - both small and large language models. I'm also an advocate of dev.to, which I consider the best site of its kind. But as a Linux user, I have to skip this latest challenge - due solely to the fact that GitHub CoPilot is a Microsoft product.


    It's hard to separate a product from its parent company’s broader philosophy. Microsoft's recent moves — the release of the telemetry-ridden Windows 11 OS and the forced obsolescence of perfectly functional hardware by ending Windows 10 support — these are practices I can't support.


    This extends to GitHub CoPilot - it's not the agentic nature that bothers me - that the application can view my files and directories and execute commands. It's that all of this isn't being processed locally - it is a verifiable fact that it is bundled and piped to Microsoft’s Azure servers.


    As Linux users, we move to distros like Debian or Arch to escape these "call-home" binaries and we reject any kind of forced ecosystem lock-in. I'm simply not installing a cloud-tethered agent that reports my terminal activity to a central server at Microsoft.


    Again, this isn't a reflection on dev.to - like many others, I enjoy these high-profile challenges. I'd just prefer that the agentic instrument was different. There is incredible work being done in the local AI space. We don't need a cloud connection to have a smart CLI agent on our PC. Open-source tools like Ollama, Llama.cpp, and local MCP (Model Context Protocol) servers prove that we can have "agentic" help while keeping 100% of our data on our own silicon.


    Zero Exfiltration: 100% of the "probing" stays on your NVMe.


    No Telemetry: No Microsoft account or HTTPS requests to Azure.


    Good Luck to those taking on this challenge - I'll be sitting this one out.


    Ben S.




    More...
Working...