AI in Software Development
What an exciting time for software developers it is. AI has had enormous impacts on software development so far. Some appreciate it as an extension and booster for writing code or even rely on it completely calling it vibe coding, others fear it as a job killer. AI in software development has different aspects. In this article I wan't to highlight some of the product categories that have come up since the launch of ChatGPT and have influenced software development. After that I wan't to make one point: Software development has radically changed.
Research on artificial intelligence has been done already for a long time, if you look it up on Wikipedia, "the field of AI research was founded at a workshop at Dartmouth College in 1956"1. So long before OpenAI released ChatGPT. But after November 2022 two magically things happened:
- It gained 100 million users by January 2023 (after two month) and "ChatGPT had become what was then the fastest-growing consumer software application in history"2.
- Everyone was talking about AI
AI product categories for software development
What did it mean for software development? I wan't to highlight the following product categories:
- AI chatbots
- AI code completion
- AI autonomous agents
There are other categories e.g. for QA, DevOps, and Documentation which I don't cover.
AI chatbots
Chatbots such as ChatGPT or later Claude became a better alternative to Google and StackOverflow for software developers because you suddenly had a bot that "understood"3 at prompt time what you meant when typing in a coding problem and printed out some code that might solve your problem. Needless to say that was a giant leap for problem solving and the beginning of the AI race.
AI code completion
LLMs trained on lots of software repositories, made it possible to offer products such as Github Copilot which is a plugin in the IDE of the software developer that auto completes the code when starting with a name or comment for a specific functionality.
AI autonomous agents
After code completion, the next step was to offer code changes from tasks inside the repository. Github Copilot Workspace let you describe a PR and then executed task autonomously inside your online repo. You then had the possibility to review the code and merge.
Then with the release of Anthropic's Claude Code, Windsurf, Cursor, Visual Studio Code Copilot in Agent Mode and OpenAI's Codex Cli it was possible to run agents on your local repository via CLI or an IDE.
Manus AI, Google AI Studio and Codex inside ChatGPT are now solutions where you can build software autonomously inside a website. You prompt what you want and the agents does the rest (Manus AI even deploys your application on it's server).
Impact on software development
How did or does it change software development? Radically. Here are my thoughts:
- It will speed up software development
- Bugs will increase
- Security issues will increase
- Costs of development will decrease
- Pressure on software developers will increase (competing against agents)
- Number of ideas implemented will rise (Entry hurdle for starting idea will be lower)
So as I said, a very interesting time for software developers 🙂 !
1: https://en.wikipedia.org/wiki/Artificial_intelligence#History
2: https://en.wikipedia.org/wiki/ChatGPT
3: GPTs are at its basic model token generator, that predict the best next token