Panel discussion: Developing in an AI world - are we all demoted to reviewers? WeAreDevelopers WebDev & AI Day March2025

AI-generated code follows a "happy path," often overlooking critical security flaws. This panel reveals how to vet AI code and avoid its hidden dangers.

#1about 2 minutes

How AI is shifting the developer role from creator to architect

AI tools are changing the developer career path, pushing engineers to focus more on high-level architecture and design rather than just writing code.

#2about 3 minutes

Addressing security and reliability in AI-generated code

AI-generated code often follows a "happy path," requiring developers to understand the underlying logic to manage security vulnerabilities and site reliability.

#3about 5 minutes

Navigating data privacy and leakage risks with LLMs

Developers must be educated on the risks of sharing proprietary code or personally identifiable information (PII) with public LLMs that use data for training.

#4about 8 minutes

Learning with AI and the risk of knowledge fossilization

While LLMs can be a learning tool similar to Stack Overflow, they risk "fossilizing" knowledge by being trained on older code and not knowing novel APIs.

#5about 5 minutes

Using private data securely with RAG and MCP

Retrieval-Augmented Generation (RAG) and Model Context Protocol (MCP) allow LLMs to use private, context-specific data from an internal data lake without public exposure.

#6about 12 minutes

Evaluating the hype and practical uses of vibe coding

Vibe coding is useful for creating disposable prototypes to test product viability but is dangerous for production due to a lack of security, scalability, and edge case handling.

#7about 5 minutes

How LLMs risk reinforcing bad coding practices

There is a significant risk that LLMs trained on large amounts of low-quality or bootcamp-level code will reinforce bad patterns and anti-patterns at scale.

#8about 6 minutes