Open Source Startup Podcast

There is genuine technical substance - micro-VM bin packing at 100k per server, snapshot-based state management enabling sub-10ms resume, the counter-intuitive point that unikernels are actually ill-suited for generic AI workloads, and the GPU timesharing problem for VMs. However, the 10ms start-time point is repeated at least a dozen times with diminishing returns, and a significant portion of the runtime is consumed by founding narrative and generic startup advice.

“you take an entire GPU, you can at most split it 16 ways, the GPU is up all the time, it takes super long to get started and then it's just attached to vm, um, forever”

“if it's going to be, as I was saying, a very generic workload where at runtime you have no idea what it's going to do like a headless browser, a full desktop environment or an agent. Then go for the Linux vm”

The most interesting contrarian move is a unikernel company's CEO openly admitting unikernels are not ideal for the hot AI-agent use case, which is counterintuitive and credible. The critique of hyperscaler compute fragmentation (Lambda vs EC2 vs Fargate) is a sharp framing. But the 'VMs are more secure than containers' argument is self-described as a former hot take now mainstream, and the entrepreneurial advice at the end is entirely generic.

“Agents and headless browsers are not ideal use cases for unicornals because you don't know what's going to run”

“I don't want to have to decide whether I need to run on Lambda or EC2 or Fargate or split my application across all of them”

Felipe is a genuine technical founder-practitioner: years of academic research, built the Linux Foundation project, named real enterprise customers (Prisma, browser use), and speaks with specific authority on virtualization internals. The credibility is real and grounded in actually having done the work, not just thought-leadership. Docked because the company is still early-stage and the conversation never surfaces business scale metrics.

“Prisma was our first customer and they were wanting to run like millions of databases in as little hardware as possible”

“I literally had discussions with the CTO of nec and he's like, listen, take it open source, make noise, then we'll bring it back in”

The episode includes useful technical specifics - 100k micro-VMs per server, 5MB unikernel vs 4GB Ubuntu VM, 9ms and 6.7ms measured response times, GPU splitting limit of 16 - and named customers and competitors (Prisma, browser use, Daytona, E2B, Microsoft Hyperlite). It is meaningfully penalised for zero business metrics: no ARR, customer count, growth rates, or funding figures beyond a vague mention of pre-seed.

“instead of running like uh, 4 gig Uboom 2 VM to run a go function, now maybe you're running a 5 meg virtual machine”

“you run a command and it says here, I'm up in 9 milliseconds...I woke up and answered you in 6.7 milliseconds”

The hosts occasionally push for useful specifics (asking for more customer examples beyond Prisma, drilling into GPU needs for VMs) and the Linux Foundation donation question is a smart angle. However, questions are routinely rambling, multi-part, and self-answering, no claim is ever challenged, and the episode closes on a textbook podcast cliché ('what's your number one piece of advice').

“So maybe talk about the new agents use cases. I know computer efficiency is certainly important, but I think people are so new to agents at this moment. I feel like how do you measure efficiency? Is it just purely to actually compute? Because agents are not really compute bound. And I think a lot of people look at sandboxes.”

“Can you talk about maybe like a couple more examples of users, uh, and like their experience before and after unicraft”