VDSG: Optimizing Multimodal AI Pipelines with Metaxy
The AI era has shifted compute toward complex multimodal pipelines, where a small input or code change can trigger expensive downstream recomputation. This talk introduces Metaxy, …
Perfecting the Art of Doing Nothing: Incremental Multimodal AI Pipelines with Metaxy
AI pipelines are now expensive enough that recomputing more than necessary is the dominant cost. Tokens and GPU hours change the economics, and agentic workflows branch and …
Docling + Metaxy: Patent Intelligence at Scale
A walkthrough at the Docling Community Office Hours of how Jubust combines Docling for document parsing with Metaxy as an incremental, field-level metadata control plane to build …
The Bitter Lesson Stops at the Lab Door
The hard part of AI is not getting a model to perform in a demo. It is connecting it to real systems, giving it usable context, and keeping it under control once it matters.
Stacked PRs and AI Worktrees
AI coding agents increase output, but they also increase pressure on code review and branch management. Stacked pull requests and isolated worktrees help keep changes small and …