Noteworthy: this is powered by a new ultra-low-latency userspace TB5 driver offering an ibverbs/RDMA interface over thunderbolt 5[1]. Non-inference HPC workloads could see a massive benefit as well.

[1]: See MLX integration PR at https://github.com/ml-explore/mlx/pull/2808

> Notably, the cluster used less than 500 watts of power

Still startlingly close to microwave oven wattage.

The other notable thing is still waiting for M5 Max and M5 Ultra which should be a big performance increase and maybe more RAM per device, should be amazing for small clusters.