[P] We made GoodSeed, a pleasant ML experiment tracker

Independent developers have launched GoodSeed v0.3.0, a new open-source machine learning experiment tracker positioned as a faster, simpler alternative to established platforms like Neptune. The tool, available via pip install and with a demo at goodseed.ai, is built with a clean, beautiful UI and core functionality for tracking metrics, configs, and system resources. A standout feature is its Neptune proxy, which allows users to directly view their existing Neptune runs within GoodSeed's interface and even migrate them to GoodSeed's local or beta remote storage, lowering the switching cost for teams.

The technical implementation includes robust monitoring plots for GPU/CPU usage, memory, and power consumption with support for both NVIDIA and AMD hardware—a notable inclusion often missing from other trackers. It handles metric visualization with zoom-based downsampling and smoothing, structured configs in an interactive table, and git status logging for experiment reproducibility. The developers are currently offering a beta remote server for online syncing with limited free access, hinting at a potential future subscription model due to server costs. As an MIT-licensed project on GitHub, GoodSeed represents a significant new entrant in the ML ops space, providing a fully-featured, open-source option that could pressure commercial incumbents on price and transparency.