Cohere buys code reviewer cubic. A second pair of eyes joins the team.

An AI that writes code is useful. An AI that calmly asks why ringing the doorbell also opens the garage sounds like a useful colleague, too. That is my imaginary example of code review: understanding how a change can affect another part of an application.

Canada’s Cohere announced on 13 November 2026 that it had acquired code-review company cubic. The team will build a secure software development business within Cohere. [1] Canadian technology publication BetaKit reported the acquisition that same day. [2]

The reviewer joins the team

cubic uses AI to review proposed code changes and flag possible bugs and vulnerabilities. Cohere wants to connect that expertise to its wider enterprise platform. According to the company, this includes environments where organisations manage their own infrastructure. [1]

AI code review was not invented that day. The news is that Cohere has brought this specialist team and its technology on board. The promised further integration is not a product improvement we have tested ourselves.

That distinction deserves its own drawer. Otherwise, every change of business card might accidentally get filed under inventions.

One announcement, two dates

The same announcement also mentions security company AISLE. But AISLE had already announced on 10 November that its acquisition was complete and that it would continue as Cohere Cyber. [3] We therefore do not count that completion as a second new event on 13 November.

AISLE focuses on finding security problems and helping to fix them; cubic focuses on reviewing software code. Cohere is bringing these activities under a broader corporate roof. That may support collaboration, but it does not, by itself, prove that programs immediately become safer.

What can readers take from this?

For teams using these tools, I would look beyond the technical features and ask what an acquisition changes in practice. Who supports the product? What agreements govern how the code is processed? And which of the announced connections are actually available?

Those are my questions about this development, not a report that anything has gone wrong with customer data. The announcements we read do not yet provide a complete practical test of the combined offering.

I find the direction interesting: investing seriously in careful reading as well as faster writing. My pencil likes to hurry. My eraser would rather it paused occasionally.