Undo Raises $37M to Fix AI Code Chaos

Jan de Vries · 2026-06-15

Cambridge-based Undo just announced a $37 million funding round to help engineering teams regain control over AI-generated code. The round was led by Elsewhere Partners, and it's a big bet on runtime context — the missing piece in modern software debugging. "We are ahead of the curve," says Undo Founder and CEO Greg Law. "Undo has spent years building deterministic, program recording technology for code failure runtime visibility, which has become absolutely essential with the rise of AI." ### Why Runtime Context Matters Now Here's the thing: AI models are great at generating code, but they're terrible at understanding it. When an AI writes code, it doesn't know what that code actually does when it runs. That's where Undo comes in. Their tech captures the complete execution history of a program — every variable, every state change — into self-contained recordings. Think of it like a black box for software, but one that actually helps you fix problems. "This investment allows us to accelerate at exactly the right moment," Law adds. "Embedding Undo into AI workflows, scaling our commercial reach, and ensuring we are an essential part of how engineering teams operate in this new, AI-first world." ### The Bigger Picture: AI Code Is Getting Messy It's not just Undo. Throughout 2026, we're seeing sustained funding activity around the operational layers of AI-assisted software development. Think software supply chain control, AI-agent security and governance, CI/CD, cloud operations, production infrastructure, and developer tooling. Companies like Cloudsmith, Geordie AI, Trent AI, CodeWords, Overmind, and Toyo have all raised money this year. Undo's round fits right in because it addresses a core enterprise concern: how do engineering teams maintain reliability, visibility, and control as AI-generated code makes systems more complex? "AI is making code unmanageable," says Elsewhere Operating Partner Rod Favaron. "It introduces code that engineers cannot understand, trust, or debug. So while AI helps them generate more code, some of it is poorly understood, poorly structured, and of questionable quality. Systems become full of unknowns, making them unstable and increasing the risk of outages, security breaches, and customer escalations." ### How Undo Actually Works Founded in 2012, Undo lets coding agents solve complex problems on complex codebases. They fill a critical gap by giving AI agents the runtime context needed to reliably diagnose software issues. By capturing complete execution history into recordings, Undo enables fully automated root cause analysis — across development, test, and production. Here's the key insight: AI = model + context. No matter how good the model is, results are limited by the quality of the context. Runtime context tells the model what the program actually did, not just what the code says. And that makes all the difference. ### Real Results That Matter Undo's benchmarks on a range of complex bugs show dramatic improvements: - The latest models identify root cause just 38% of the time without Undo. With Undo's runtime context, that jumps to 92%. - When models can solve a problem without Undo, using Undo means fewer tokens are needed. - Customers report completing root cause analysis 100 times faster with Undo than before. "Quality is extremely important at Palo Alto Networks, and we cannot afford to rely on guesswork," says Suresh Sangiah, Senior VP Engineering at Palo Alto Networks. "The hardest — and costliest — bugs in multi-million-line codebases live in runtime state and are not captured by logs or other solutions. Undo provides the visibility needed to catch and correct errors before they become an operational problem for our customers." ### What This Means for Engineering Teams If you're managing a team that's adopting AI-assisted coding, Undo's approach is worth watching. The promise is simple: AI writes more code, but Undo makes sure that code actually works. That's a value proposition that resonates in an era where system complexity is growing faster than our ability to understand it. For US-based engineering leaders, the takeaway is clear. As AI-generated code becomes more common, tools that provide runtime visibility aren't just nice-to-haves — they're becoming essential infrastructure. And with $37 million in new funding, Undo is positioning itself to be exactly that.