The real bottleneck in software development has moved. It is no longer writing code. It is reviewing it.
That is the blunt assessment from Qodo CEO Itamar Friedman, and it is the problem his company’s latest platform update is designed to solve. Version 2.8 of the Qodo platform, announced this week, targets a specific pain point: the growing gap between how much code gets generated and how much a human team can actually check.
The numbers are not in the release, but the logic is straightforward. AI coding tools let developers produce more lines of code, faster. That means more pull requests.
More changes to review. More chances for something to break.
Human reviewers, however, do not get faster. They stay the same. The queue grows.
Qodo’s answer is a set of features built around automation and pattern recognition. The most notable is a custom rules miner. It scans existing codebase behavior and pull request history to discover the coding patterns a team actually uses.
It then turns those patterns into structured, enforceable rules. For a team of developers sharing a project, the pitch is simple: less time arguing over tabs versus spaces or bracket placement, more time building.
The platform also adds a capability to discover what Qodo calls “AI skills.” These contain code review instructions, coding standards, and engineering best practices drawn from across multiple repositories. Those skills are surfaced in a central portal. DevOps teams can manage them from one place and assess their impact on the software engineering workflow.
For teams scattered across cities like Manila, Singapore, or Jakarta, that centralization could mean fewer late-night emergencies over broken builds. Friedman described these capabilities as extending an agentic AI platform for governing code.
The underlying technology is graph-based. It tracks the relationships that exist between pieces of code. When a pull request modifies a shared dependency, the agent reads the affected repositories.
It surfaces impact findings before the pull request is merged. What that looks like in practice is specific. The system flags issues like function signature violations.
It catches contract breaks between application programming interfaces. It spots changes to schemas.
It detects infrastructure drift. These are the kinds of problems that slip through in a fast-moving development cycle and cause hours of debugging later. The implications for engineering teams are practical.
The bottleneck has shifted, and the tools are shifting to meet it. The question is whether they can keep pace.
For now, Qodo is betting that graph-based code governance and automated rule enforcement are the right approach. The alternative, Friedman implied, is a system where human reviewers are simply overwhelmed by the volume of code coming at them. The platform is live.
Teams can start using it now. Whether it actually cuts down on those late-night firefights over broken builds remains to be seen. But the logic of the problem is hard to argue with.
If code gets easier to write, review has to get easier too. Or everything slows down.
