Since the launch of public access to ChatGPT, much has been made of the impact that generative artificial intelligence will have on the profession of software development.

However, I think that there will be important milestones before this impact is profound. The two milestones are:

AI that can debug code as an agent

If AI can run as a agent on a development server and also has permission to edit application code, this will “close the loop” as a human developer will not need to be involved in order for the AI agent to iteratively debug software.

Developers will need to focus more on system architecture, defining requirements and project management — allowing AI to handle the lower-level coding implementation details.

AI that can translate business requirements into code

AI has proved itself to be quite capable with snippets at code, but has yet not the context length to absorb entire repositories. It is yet to be proved how well AI will be able to ingest an entire repository and then understand the business outputs of the code. Until then, a developer will still be required to translate business requirements into code requirements.

Summary

The software development profession stands at the precipice of an AI-driven paradigm shift. Those who upskill and pivot to developing AI-augmented software will survive in the long term.