The triple double. We deliver two Firecracker microVMs for $2/month with a two-person team and we're wildly profitable doing it. Every wasted cycle is margin we're leaving on the table. If that constraint excites you, keep reading.
You know that old crotchety guy in the PR comments? The one who asks "have you profiled it?" on every diff. Who thinks malloc is a code smell. Who says most software is just badly disguised memcpy. Who won't shut up about locality.
That's us. We deliver two Firecracker VMs for $2/month — one for the customer, one given free to a developer in Latin America — and we're profitable on every single account. Not VC-subsidized. Not "we'll figure out monetization later." Profitable. Right now.
You don't hit those margins with sloppy code. Every wasted cycle is money. Every unnecessary allocation is a VM we can't pack onto the box. Every TLB flush we don't need is latency a customer feels. And the best code is the code you didn't write — sometimes the win is a kernel flag, a TCP tuning parameter, a cgroup limit you didn't know existed. This is infrastructure where the crotchety guy's obsessions aren't neurotic — they're the entire business model.
We're a small team in Paraguay. We ship fast, we don't do meetings, and we care more about your perf intuition than your resume.
Go and Rust on bare metal Linux. Firecracker VMs with custom kernels. NVMe swap orchestration. KSM tuning. Custom DNS. JuiceFS for distributed storage. Caddy for TLS termination. Postgres. No Kubernetes. No Docker in production. No YAML pipelines. Just code that talks to the hardware. We build the cloud how it was supposed to be — we don't run on the cloud.
The constraint is real: two VMs, $2/month, profitable. That means every architectural decision matters. You don't get to throw RAM at the problem. You don't get to "just add another node." You get to be clever, and the business depends on it.
If you've ever wished your job involved more perf stat and less Jira — or Linear — this is it.
Two-person team, massive surface area. One day you're tuning kernel swap parameters, the next you're writing landing page copy, the next you're sourcing cheaper bare metal in new regions. Marketing, kernels, billing, infra, Stripe webhooks, cold emails to datacenter sales reps, buying old systems on eBay — whatever moves the needle that week. You're not a specialist. You're the third 2.
RemoteSend an email to jobs@lobstertank.me with something you've hacked. Something that wasn't supposed to work. A thing you broke spectacularly and learned from. A gist, a blog post, a war story — fuck it, even a spectacular fail. Tell us a story, not a CV.
No take-home assignments. No whiteboard puzzles. We'll just talk about systems.