AI can generate code - but it still needs human review
While AI can generate a significant amount of code, and do so far faster than humans can, that doesn't mean that it can be blindly trusted. In this episode, I discuss a study from 2025 which investigates how code security tends to change (and not always positively) when it's iteratively worked on by AI tools.
It's an extremely insightful study, one that clearly shows that AI-generated code needs to be regularly reviewed and that humans still need to be in the loop of software development. So, for a rolling (and somewhat meandering) discussion of the study, join me in this episode, and hear what three researchers have to say about AI code generation.
Links from the episode
It's an extremely insightful study, one that clearly shows that AI-generated code needs to be regularly reviewed and that humans still need to be in the loop of software development. So, for a rolling (and somewhat meandering) discussion of the study, join me in this episode, and hear what three researchers have to say about AI code generation.
Links from the episode
Want to grow your knowledge further?
Check out the books below, where you can grow your web development skills in a very focused, very hands-on way.
Check out the books below, where you can grow your web development skills in a very focused, very hands-on way.
Hosted and produced in Bundaberg by: Matthew Setter.
Website: https://matthewsetter.com.
Follow me on: LinkedIn.
If you like the podcast, and want to support it, how about buying me a coffee?
★ Support this podcast on Patreon ★
Follow me on: LinkedIn.
If you like the podcast, and want to support it, how about buying me a coffee?
