Greg Kurtzer launches OpenWaldo framework for open source AI
Greg Kurtzer launched OpenWaldo, a community-backed corpus of AI training data that includes an AI bill of materials. The project argues that true open source AI requires access to training data and recipes, not just model weights.
Greg Kurtzer, founder and CEO of CIQ and creator of CentOS and Rocky Linux, launched OpenWaldo, a community-backed corpus of AI training data that includes an AI bill of materials. The name is an acronym for Open Weights, Open Artifacts, Open Licenses, Open Data, and Open Origins. Kurtzer argues that current 'open source' AI models are misleading because they typically only release binary weights under licenses like Apache 2 or MIT, without providing the source data or training recipes required to reproduce the artifact. He asserts that true open source AI requires all components to be available for auditing and rebuilding.
OpenWaldo provides a 'compose file' that allows users to replicate the exact training process, though outcomes may vary due to non-deterministic systems. Kurtzer distinguishes this from frontier models, stating the project aims to build a foundation model and community infrastructure rather than competing with major labs. He cites Tim O'Reilly’s concept of 'architecture of participation' as a guiding principle, emphasizing that open source AI must include community collaboration and tooling, not just open weights. The project is currently in early stages, with Kurtzer seeking community help for corpus development, tooling, and marketing.