As AI coding tools generate more pull requests than ever, code review has become the new bottleneck in software delivery. Human reviewers struggle to keep up with the volume, and generic AI review bots often produce noisy, low-value comments that teams learn to ignore. cubic, a Y Combinator (X25) startup founded in 2025 and based in London, set out to build an AI-native code reviewer specifically designed for complex codebases and the AI era of software development.
cubic's reviewer analyzes proposed changes with deep context about the surrounding repository, helping it understand how an edit fits into the broader system rather than evaluating diffs in isolation. This context-awareness lets cubic catch genuine bugs, logic errors, and downstream impacts while keeping false positives low, addressing the signal-to-noise problem that undermines many automated review tools. The goal is to give teams reviews they actually trust, so they can merge confidently and quickly.
The company positions cubic as Cursor for code review, drawing an analogy to the popular AI code editor: just as Cursor reimagined writing code with AI, cubic aims to reimagine reviewing it. Early adopters include well-known developer-focused companies such as Cal.com and n8n, and cubic reports that its platform can meaningfully increase pull-request merge throughput, with figures around a 40 percent improvement cited for some teams. The company has continued to iterate quickly, shipping a cubic 2.0 release touting substantial gains in accuracy and speed.
cubic is early-stage, having raised around $500K from investors including 468 Capital, Batch Ventures, and Y Combinator. Founded by Paul Sanglé-Ferrière, the small London team is focused on proving that AI can make code review faster and more reliable, turning a painful gate into an accelerator for teams shipping AI-assisted software.