The era of "observability for everything" is about to hit a wall built by developers who just want their AI-generated code to stop breaking in production.
The Summary
- Sazabi raised $8 million seed led by J2 Ventures, Village Global, and Y Combinator to build observability tooling specifically for teams shipping AI-assisted code
- The platform strips away dashboards and metrics in favor of one data type: logs, the time-stamped record of what actually happened when your agent-written function exploded
- Over 60 angels from Cursor, OpenAI, Anthropic, and Replit backed the round, which tells you who's feeling the pain of debugging AI-generated code at scale
The Signal
Former Brex engineer Sherwood Callaway built Sazabi after watching engineering teams drown in observability data that didn't help them ship faster. The platform targets startups using AI coding tools like Cursor and Claude Code, which means it's solving a problem that didn't exist two years ago: how do you monitor code you didn't fully write?
The thesis is simple. When production breaks, engineers go straight to logs. Not dashboards. Not metrics. Not trace visualizations. Just the raw chronological record of what the system did before it fell over.
"Whenever there was a problem in production, I would reach for a log stream. This is the type of data that felt most intuitive to me."
Legacy observability platforms like Datadog and Grafana were built for a world where humans wrote every line and understood system behavior. AI-assisted development breaks that model. Code ships faster, engineering teams are leaner, and the surface area for unexpected behavior expands exponentially. You need tooling that matches the velocity and chaos of agent-augmented workflows.
The company came out of YC's spring 2026 batch and immediately pulled in angels from the companies building the AI coding tools themselves. That investor list matters. When people from Cursor, Anthropic, and Replit write checks, they're signaling that the debugging problem for AI-generated code is real and unsolved.
Sazabi's pricing model reflects the new economics of AI-native infrastructure:
- Free tier for experimentation
- Credit-based system tied to tokens used and log volume
- Built for teams that scale unpredictably because their agents do
Callaway calls the internal mission "Operation Waterloo" because Datadog is French and he wants to beat them. That's not just founder bravado. It's a bet that the next generation of observability doesn't layer on complexity. It strips it away for teams moving too fast to babysit dashboards.
The Implication
If you're building with AI coding assistants, you're probably already feeling this. Your team ships faster but spends more time debugging production issues you didn't anticipate. The old tools don't fit the new workflow. Watch for more infrastructure plays purpose-built for agent-augmented engineering teams. The companies that win here won't add features. They'll subtract friction for developers who just need to know what broke and why, five minutes ago.