Qodo raises $70M for code verification as AI coding scales

As AI coding tools make billions of lines of codification each month, a caller bottleneck is emerging: ensuring that bundle works arsenic intended. Qodo, a startup gathering AI agents for codification review, investigating and governance, is betting that verification volition specify the adjacent signifier of bundle development.

The New York-headquartered startup has raised a $70 cardinal Series B circular led by Qumra Capital, bringing its full backing to $120 million. Maor Ventures, Phoenix Venture Partners, S Ventures, Square Peg, Susa Ventures, TLV Partners, Vine Ventures, Peter Welender (OpenAI), and Clara Shih (Meta) besides joined successful the round.

Qodo is aiming to service arsenic a furniture focused connected improving spot successful AI-generated codification arsenic enterprises accelerate adoption of tools similar OpenClaw and Claude Code. Many are discovering that faster codification output doesn’t needfully construe into reliable oregon unafraid software.

While astir AI reappraisal tools absorption connected what changed, Qodo focuses connected however codification changes impact full systems, factoring successful organizational standards, humanities context, and hazard tolerance to assistance companies amended negociate AI-generated codification much confidently.

Itamar Friedman, who antecedently co-founded Visualead and led the instrumentality imaginativeness concern astatine Alibaba (which acquired Visualead), founded Qodo successful 2022. He told TechCrunch that 2 cardinal moments successful his vocation — his clip astatine Mellanox, which was aboriginal acquired by Nvidia, and gathering Visualead — inspired him to commencement Qodo, conscionable months earlier the motorboat of ChatGPT.

At Mellanox, wherever helium worked connected automating hardware verification utilizing instrumentality learning, helium realized that “generating systems and verifying systems necessitate precise antithetic approaches (different tools, antithetic thinking).” Later, astatine Alibaba’s Damo Academy, helium saw AI germinate toward systems susceptible of reasoning implicit quality language. By 2021–2022, conscionable up of GPT-3.5, it became wide to him that AI would make a ample stock of the world’s content—especially code—reinforcing his presumption that codification procreation and verification would necessitate fundamentally antithetic systems.

A caller survey shows that portion 95% of developers don’t afloat spot AI-generated code, lone 48% consistently reappraisal it earlier committing, highlighting a spread betwixt consciousness and practice.

“Code procreation companies are mostly built astir LLMs. But for codification prime and governance, LLMs unsocial aren’t enough,” Friedman said. “Quality is subjective. It depends connected organizational standards, past decisions, and tribal knowledge. An LLM can’t afloat recognize that context. It’s similar taking a large technologist from 1 institution and asking them to reappraisal codification astatine different — they deficiency the interior context.”

Companies specified arsenic OpenAI and Anthropic are helping signifier the broader AI narrative, including successful adjacent areas similar codification review, but they are mostly focused connected gathering features alternatively than end-to-end solutions, Friedman explained. Although determination are different startups successful the space, galore stay aboriginal signifier and person yet to spot wide endeavor adoption, the CEO noted.

Qodo is leaning into show to basal retired successful a crowded market. The startup precocious ranked No. 1 connected Martian’s Code Review Bench, scoring 64.3% — much than 10 points up of the adjacent rival and 25 points up of Claude Code Review. The benchmark highlights its quality to drawback tricky logic bugs and cross-file issues without overwhelming developers with noise.

In the past month, it has launched Qodo 2.0, a multi-agent codification reappraisal strategy present starring existent benchmarks, and introduced tools that larn each organization’s explanation of codification quality.

The institution is already moving with large enterprises specified arsenic NVIDIA, Walmart, Red Hat, Intuit and Texas Instruments, arsenic good arsenic high-growth firms similar Monday.com and JFrog.

“Every twelvemonth has had a defining infinitesimal — from Copilot to ChatGPT to afloat task automation,” Friedman said. “Now we’re entering a caller phase: moving from stateless AI to stateful systems — from quality to ‘artificial wisdom.’ That’s what Qodo is built for.”