Generative AI is boosting software development productivity, but growing volumes of AI-written code can also raise review, fixing and maintenance costs, an analysis said. Deepak Anupal, co-founder and CTO of WaveMaker, argued the key is not how to inspect generated code but how to generate less. He proposed an “assembly model” that prioritises reusing verified components and libraries, reducing repeated checks and limiting new code to what is essential.

Security firms are publishing early experiences of Anthropic’s AI model Mythos, which was first shared on a limited basis with some companies, with Cloudflare standing out. Cloudflare said Mythos is a tool on a different level from earlier general-purpose frontier models, citing attack-chain building and proof generation. Separately, Anthropic is allowing Mythos users to share cyber threat information with other companies exposed to similar vulnerabilities.