Scale Kubernetes to Zero, Wake it up Instantly

New York, New York
📢 Sandboxes, get your sandboxes! 📢 We know you’re not happy with your current sandbox provider, and we’ve got the solution! Check out agx.so - Loophole’s sandbox service. Built on our internal hypervisor, substrate. And yes, it runs locally on your laptop
I may be a bit late to the “time to interactive” agent sandbox party, but I wanted to make sure our hypervisor absolutely MOGGED everyone else. 10x faster than Firecracker and 425x faster than Cloud Hypervisor. agx.so is the best agent infra platform, period.
2
1
11
2,690
Loophole Labs retweeted
LiveKit has acquired @LoopholeLabs, an infrastructure company whose technology will enhance our platform for building voice, video, and physical AI agents. With their team, we move closer to offering customers essentially unlimited concurrent agents. Welcome to LiveKit! fortune.com/press-releases/l…
3
14
81
3,356
Loophole Labs retweeted
Remote GPUS :D
1. Reverse engineer the CUDA API 2. Because why not? 3. ... 4. Profit! Or, what exactly is gtap.ai all about featuring @confusedqubit piped.video/qS1BEzkOqMU
1
3
1,946
1. Reverse engineer the CUDA API 2. Because why not? 3. ... 4. Profit! Or, what exactly is gtap.ai all about featuring @confusedqubit piped.video/qS1BEzkOqMU
2
9
2,760
Loophole Labs retweeted
Spot is cheap until the node gets reclaimed and your in-memory state vanishes. Architect catches the notice and moves the live container before the node gets reclaimed, memory and all. Zero downtime, no data loss. Read more: loopholelabs.io/blog/statefu…
1
2
218
Many scale-to-zero products have an ugly reality: they kill your pods. So every wake is a full restart. Schedule, pull, init, boot often resulting in 90+ seconds of nothing. With Architect pods wake from a checkpoint. True scale to zero. loopholelabs.io/blog/scale-t…
1
9
6,468
Loophole Labs retweeted
Full website and release coming soon! This is all part of our new sandbox product. Will work both locally (Mac, Linux, Windows WSL) and in the cloud seamlessly - with *all* the features, like network MITM, secrets injection, etc. And some truly bonkers stuff for agent platforms
1
2
6
299
Loophole Labs retweeted
Not a bad cold boot time using FC's own benchmarking methodology. Substrate is @LoopholeLabs's hypervisor, designed for the ground up for agents and agent platforms. If you're building an agent platform, you deserve better than just another Firecracker/Cloud-Hypervisor wrapper
1
2
20
1,964
Our customers hit the hard edges of CRIU. So we faced with an age old question: Can we solve these problems in the existing project, or do we need to create a new one? AI didn't make the choice but it did change the outcome. We're rewriting CRIU in Zig: loopholelabs.io/blog/rewriti…
1
4
476
Loophole Labs retweeted
Been benchmarking the AGX cloud scheduler this week! 5 million sandboxes *started* in under a minute. Not scheduled, running. We don’t have to find the optimal bin packing at schedule time because we can just live migrate later! Love using the tools we’ve built @LoopholeLabs
2
1
17
1,748
Ever want to do GPU heavy work on a system without a GPU? Perhaps Ollama on a MacBook. What if you could use a GPU over the wire to do it? Now you can with GTAP, read more in our blog: loopholelabs.io/blog/ollama-…
1
312
Loophole Labs retweeted
Replying to @damian_b
Our sandboxing solution. Takes all of @LoopholeLabs’s magical tech and puts it into the *best* sandboxing experience possible. Local first, instant snapshots, and we’ve even gone as far as to make native macOS snapshots work on Linux kvm and back. Makes it trivial to move sandboxes around.
1
8
431
Loophole Labs retweeted
By the way, if you’ve been enjoying my hot takes on cloud infra, ebpf, live migration, and agent sandboxes @LoopholeLabs is hiring! loopholelabs.io/careers Come build the future of infrastructure with me!
2
5
35
3,292
Loophole Labs retweeted
Fun little benchmark: AGX can cold-restore (ie. fresh host, no cached data except for the 16MB linux kernel binary) a 16GB VM with a 128GB disk in *84ms*. The techniques we use here are in the same family as the ones we use for architect.io at @LoopholeLabs
1
2
6
649
Loophole Labs retweeted
We’ve been solving GPU supply chains @LoopholeLabs so you can get GPUs whenever, wherever, for however long you need via gtap.ai - coming soon! DM me for early access!
checked in this morning, all 8x H200/B200 sold out. prime intellect, runpod, lium, verda, lambda. so where do you go to get GPUs anymore? do i need to build a GPU-sniper to get them?
2
10
2,494
Loophole Labs retweeted
Business Cards // @LoopholeLabs └ loophole-labs-cards.png +-+-+-+-+-+-+-+-+-+-+-+ Featuring: {∞} Loophole Labs + /~\ Architect The Wizard aka Archie (ASCII) Team faces (^_^) <•‿•> {°…°} |#‿#| Commit Mono (500) PMS 2665, 165 +-+-+-+-+-+-+-+-+-+-+-+
1
2
352