AI writes so much code that another AI must begin reviewing it

Artificial intelligence has reduced the time needed to produce software. It has not reduced the time needed to trust it by the same proportion.

CodeRabbit has raised $143 million in a round valuing the company at $1.5 billion. Its business is not primarily writing more code, but reviewing changes, explaining their effects and identifying vulnerabilities or maintenance risks.

The platform performs more than two million reviews each week and reports over 17,000 customers, including Nvidia, BMW, JFrog, Trivago, Adyen and Indeed. The scale suggests verification is becoming a market of its own.

A conventional code review compares a change with the existing system. It looks for errors, duplication, security problems and decisions that will make the product harder to maintain. It also transfers knowledge between developers.

When an agent can modify dozens of files in minutes, volume exceeds available attention. Reviewing line by line cancels part of the speed gained; accepting everything without checking moves the cost into future failures.

CodeRabbit proposes an independent layer. It reads the change, relates files, summarises intent, flags risks and prioritises what deserves human attention. The company calls its expansion agentic change management.

i

i

i

i

←01 / 04→

Independent does not mean infallible. A model can approve an error, create irrelevant alerts or miss an internal rule. Its useful role is narrowing the search space and preserving traceability, not eliminating responsibility.

The analogy with architecture is direct. Generative tools can produce options, details or documents quickly, but the profession needs systems that test codes, coordination, geometry and changes between versions.

If generating ten solutions costs almost the same as one, scarcity moves to the ability to compare them. Professional value shifts from producing volume to building criteria, tests and limits.

The funding also answers cyber risk. More capable models allow more sophisticated attacks, and companies need to inspect more software. CodeRabbit plans to spend over $10 million next year to keep review tools free for open-source projects.

Growth opens another question: if the same industry uses AI to produce and validate, what happens when both systems share biases or lack the same context? Separate suppliers and deterministic tests will form part of real governance.

The fact. $143 million in funding, a $1.5 billion valuation and two million weekly reviews. The editorial reading is that abundant production turns verification into infrastructure. The creative future will not have less control; it will need better-designed control.

i