NxCode Raises New Funding to Build Verification Layer for AI-Generated Software

NxCode, an AI development platform that turns plain-English descriptions into deployable software, has raised new funding. The capital will go toward the company’s multi-agent engineering system and the verification layer that reviews what those agents produce.

The company says that as code generation becomes more widely available, it sees judgment — deciding what is worth building, recognizing which failures are acceptable and which are not, and knowing when a system should stop and ask instead of proceeding — as the more durable differentiator.

“Anyone building software has to think about where the world is heading, and whatever can be free will end up free,” said Zirong Chi, NxCode’s founder. “What is left is judgment — deciding what is worth building, and what it costs when you get that wrong.”

NxCode’s platform is aimed at users without engineering backgrounds who are building software themselves. The company points to this as a reason its verification layer matters: failures such as a broken account-recovery flow, a database without a safe migration path, a payment flow that mishandles an edge case, or an agent surfacing information it should not have held can be difficult for non-technical users to catch on their own.

NxCode’s approach is structural: the platform assigns planning, architecture, coding, testing and deployment to specialized agents. The company says this verification layer includes what is checked automatically, which risks are explained in plain language, when the system declines to continue, and how a user can inspect, export or roll back what the agents produced.

“Most people’s way of thinking about a problem isn’t actually aimed at solving it,” Chi said. “I have a lot of respect for anyone willing to face one head-on. That is why I left a comfortable, retired life in Japan to start this.”

The company said it expects to be measured on operational metrics rather than launch volume: how many projects reach production, how many are still running months later, how often users need human help, which errors recur, and whether customers can move their code elsewhere without penalty.