 So auto retrieve, we've mentioned a couple of times. This is one of the stop gaps that we're putting in place so that in the short term, we can make content that's in Filecoin accessible to IPFS, to gateways and just more generally bridge some of the protocol gaps that we've got at the moment. It also serves a secondary purpose, which is it gives us a lot more view into the state of retrievals and lets us work with data programs to sort of help set up the right incentives to encourage storage providers to ramp up on their retrieval bandwidth and their infrastructure so that they can serve the amounts of retrievals that we're expecting to keep growing. So this is running, we've recently switched it to a Kubernetes deployment that we can keep running pretty stably. We're working through some ongoing resource management stuff so that it not only is running, but also serving at high quality. You can see some gaps in the success failure rate where it runs out of memory currently. All this work is thanks to Elijah on the OuterCore team and Kyle on the Bedrock team, but more generally what this is going to mean is that when you go to ipfs.io, what will happen is that will go back to the big IPFS node that is that gateway. It will be peered and so it's bit swap requests. We'll talk to its peers and one of the peers will be this auto retrieve node which looks like an IPFS node that is just sort of in the IPFS node. Right now, you need to be peered. What that means is it's serving currently IPFS nodes that are in the DHT server ring because it automatically connects to them. But if you're another IPFS client, you're not getting the full benefit quite yet because you won't necessarily be connected. Those bits off requests will then be seen by Outer Retreat who will ask to store the index indexing node for those sids. When those sids are found from a storage provider on Filecoin, it will then make a graph sync request to pull that content locally into its own cache and then we'll say that it has those blocks and be able to respond to them over bits well. So it acts like a block cache. It keeps a relatively large order of tens to hundreds of gigs of blocks that it knows about in cache that it's pulled from storage fighters. But the thought is that these are transient. We can eventually have them running in the same regions as gateway instances and just generally use this as a short-term over the next month's way to bridge until we get some vertical upgrades. I will leave it there. There's an Outer Retreat channel at Filecoin Slack.