AI has changed the way software engineering works - including introducing new risks -getty

‍

As AI coding tools take on more of the actual work of software engineering, companies are discovering a messy problem: nobody really knows what's going to break until it already has. Google's DORA report found deployment stability has actually gotten worse since AI coding took off, down 7.2%, and a third of developers admit they don't fully trust AI-written code. Causal Dynamics Labs, an $8M seed-funded startup, is betting that the root cause is context. Their research found AI agents spend more than half their time just searching for files, essentially wandering around a codebase trying to figure out how things fit together before they can do anything useful. Their fix is to build a "digital twin" of a company's software, a living map of how everything connects, so changes can be tested against reality before they go live. They're already piloting with 40+ Fortune 500 companies and converting roughly a quarter into paying customers.