According to gdb, OpenAI reveals MRC and supercomputer networking engineering, tackling networking, scheduling, storage, and reliability at planet scale.

According to @gdb, OpenAI launches MRC with AMD, Broadcom, Intel, Microsoft, NVIDIA to speed training and cut GPU waste, now running in production.

According to @OpenAI, experts detail Multipath Reliable Connection to sync record chip counts and move data efficiently for AI supercomputers.

According to OpenAI... MRC, built with AMD, Broadcom, Intel, Microsoft, and NVIDIA, improves large AI training throughput and reliability with less GPU waste.