AI Weekly Malaysia

Hugging Face ran a 19-day hackathon (July 15–Aug 2, 2026) where 1,200+ participants used coding agents (Claude Code, Codex, Cursor, OpenResearch's orx) to reproduce ICML 2026 papers claim by claim, producing 6,816 Trackio logbooks covering 2,226 of the conference's 6,352 accepted papers (~a third). ICML 2026 saw 23,918 submissions and 6,352 acceptances, roughly double the prior year, partly attributed to AI agents accelerating experiment cycles and writing.

Why: If you build or rely on ML research outputs, this signals that a meaningful fraction of top-conference papers may not have been rigorously verified by reviewers—one spotlight paper's proofs went unchecked despite strong scores. Coding agents can now attempt full reproductions in an afternoon that would cost a human reviewer a weekend, so teams shipping ML should consider running agent-based reproduction checks on papers they depend on rather than trusting acceptance as a quality signal.

AWS authors walk through a continuous data loop using Strands Robots (Apache 2.0 SDK), LeRobot's on-disk dataset format (90,000+ datasets, 8,000+ publishers on HF Hub), and Hugging Face Storage Buckets (mutable, non-versioned, Xet-backed object storage announced March 2026) to record robot demonstrations, train policies on growing datasets, and deploy back to hardware like the SO-100/SO-101 arms. The core problem they address is that running this loop daily causes repeated full-dataset transfers to GPUs and redundant byte costs, which the Storage Bucket layer sitting in the hf:// namespace is meant to mitigate as a working staging layer between recording and training.

Why: If you are building robotics learning pipelines with LeRobot-compatible datasets, HF Storage Buckets give you a mutable staging layer that avoids re-copying the entire dataset to GPUs on every training run—a concrete cost and workflow decision. For everyone else, this is a niche robotics tooling walkthrough that doesn't require any change to what you ship.