All videos
0:00 / 0:00
research

AI Dev 26 x SF | Tom Howlett: Can LLMs Generate Enterprise Quality Code?

DeepLearning.AI16 June 2026Watch on YouTube

What you'll learn

  • LLMs vary significantly in code quality: some models produce more than 2x the issues of others.
  • Static analysis can make AI agents safer and more reliable for enterprise applications.
  • AI-generated code must be evaluated on reliability, maintainability and security, not just speed.
  • Developers can integrate deterministic static analysis with AI agents to maintain enterprise-level quality without sacrificing productivity.

Frequently asked questions

How many LLMs were tested in Sonar's benchmark?
Sonar tested more than 35 of the latest and highest-performing large language models and compared them on code quality.
What is the biggest difference in code quality between the tested LLMs?
Some models produce more than 2x the issues of other models, showing that model selection significantly impacts code quality.
How can developers protect AI agents from their weaknesses?
Developers can integrate static analysis with AI agents, allowing deterministic checks to validate generated code quality.
Can you maintain AI productivity while ensuring enterprise-level code quality?
Yes, by integrating deterministic static analysis with AI agents you can maintain enterprise-level quality without sacrificing AI productivity gains.

Topics

llm code generationai agentscode qualityenterprise aibenchmark

Description from the channel

We all know how fast it is to create an app with modern AI agents but how do we ensure the code is reliable, maintainable and secure enough to be used by enterprises? In this talk, Sonar's Tom Howlett shared a benchmark from their testing of 35 (and growing) of the latest and highest-performing large language models and showed how they compare not just on task completion but on the quality of the code they create. Attendees saw that models are not the same and that some produce more than 2x the issues of others. Once you understand your model how do you protect yourself from its weaknesses? Tom demonstrated how devs can integrate AI Agents with deterministic static analysis to ensure enterprise-level quality without killing the AI productivity gains.