GGML – AI at the Edge (ggml.ai)

Article summary

GGML is a tensor library for machine learning that enables large models and high performance on commodity hardware. It is used by projects such as llama.cpp and whisper.cpp. The library is designed to be simple, with a small codebase and no third-party dependencies. The company behind GGML, ggml.ai, was founded in 2023 and acquired by Hugging Face in 2026.

Main themes

• AI at the edge
• Open-source project management
• Harassment and hate speech
• Project leadership and stewardship
• Forking and licensing
• Community dynamics and conflict resolution

What commenters say

• Ejecting controversial contributors can be good stewardship, but it depends on the context and motivations.
• Some contributors were harassed and humiliated, and the project leader's response was inadequate.
• The project's leader has the right to eject anyone who creates drama or controversy, regardless of their technical skills.
• Forking the project is a viable option for those who disagree with the direction or leadership.
• The project's focus should be on building things, not on politics or drama.
• Investigating and addressing harassment and hate speech is essential, rather than just ejecting everyone involved.
• The MIT license allows for forking and modification, which can be a solution for those who disagree with the project's direction.
• Ejecting both the harasser and the victim can be seen as laziness, rather than a genuine attempt to resolve the issue.