AI Agents Need Computers: 74% MoM Growth, 850K/Day Runs, & New Agent Cloud — Ivan Burazin, Daytona

| Podcasts | May 21, 2026 | 4.3 Thousand views | 1:11:40

TL;DR

Daytona CEO Ivan Burazin explains their pivot from human developer environments to 'composable computers' for AI agents, revealing how bare-metal, stateful infrastructure (not ephemeral VMs) unlocked 74% month-over-month growth and massive enterprise demand.

🔄 The Hard Pivot to Agent Infrastructure 3 insights

Origin in Code Anywhere

Burazin and his co-founder previously built Code Anywhere, the first browser-based IDE, giving them deep infrastructure experience in virtualization and custom schedulers later applied to Daytona.

OpenDevin sparked the pivot

In late 2023, after demoing OpenDevin (now OpenHands), the team realized agents needed compute infrastructure fundamentally different from human developers, prompting a complete product pivot in January 2024.

New Year's Eve MVP

Burazin built the first prototype over New Year's Eve 2023 and despite being called 'absolute garbage' by his CTO, the concept was sound and rebuilt in two weeks, leading to immediate customer demand.

Bare-Metal Architecture Advantage 3 insights

Stateful 'laptops' vs ephemeral sandboxes

Unlike typical serverless sandboxes that preempt and destroy state, Daytona treats agent computers like human laptops—persistent, resumable, and stateful—allowing agents to pause and resume without losing context.

Bare metal speed without virtualization overhead

Running directly on bare metal with a custom scheduler rather than VMs or Firecracker provides local NVMe IOPS with zero network latency, making sandbox startup significantly faster than cloud VM alternatives.

Composable computer configurations

The platform offers API-driven composable computers enabling agents to access varying resource combinations—specific CPU counts, RAM, disk, GPUs, and operating systems—tailored to specific tasks rather than one-size-fits-all containers.

📈 Explosive Product-Market Fit 3 insights

74% month-over-month growth

Since the pivot, Daytona has experienced organic PLG-driven growth of 74% MoM, processing 850,000 runs daily, with both individual developers and enterprise customers adopting rapidly.

Customers demanding immediate access

After demoing the rebuilt product to skeptics who had rejected the earlier version, prospects began calling within 24 hours demanding API keys immediately—an unprecedented signal of market need.

Market size thesis

Burazin argues that while the market for human developer tools is large, the market for every AI agent that will ever exist represents a fundamentally larger opportunity requiring specialized infrastructure.

Bottom Line

AI agents require persistent, stateful, bare-metal compute environments that function like laptops rather than ephemeral serverless functions, and infrastructure providers offering these 'composable computers' are capturing the foundational layer of the emerging agent economy.

More from Latent Space

View all
🔬 "The Most Innovative Diffusion Research Is Happening in Drug Discovery, Not Image Generation"
1:48:40
Latent Space Latent Space

🔬 "The Most Innovative Diffusion Research Is Happening in Drug Discovery, Not Image Generation"

Evan Fineberg and Sergey Udov of Genesis Molecular AI discuss how diffusion models have pivoted from image generation to drive breakthroughs in 3D protein structure prediction. They detail how their Pearl model applies LLM-style scaling strategies—including synthetic physics-based training data and inference-time 'thinking'—to solve the historically intractable challenge of predicting how small molecules bind to proteins.

6 days ago · 7 points
AI Security After Codex and Claude Code — Zico Kolter & Matt Fredrikson, Gray Swan
1:07:31
Latent Space Latent Space

AI Security After Codex and Claude Code — Zico Kolter & Matt Fredrikson, Gray Swan

Gray Swan co-founders Zico Kolter and Matt Fredrikson explain why AI systems require a fundamentally different security approach than traditional software, highlighting how their automated red teaming system 'Shade' has begun to outperform human experts at finding model vulnerabilities. They emphasize the urgent need to treat AI agents as inherently untrusted entities capable of correlated failures across the software ecosystem.

15 days ago · 8 points