I have been heavily using AI agents recently. In my office I'm using GitHub copilot and in my personal laptop I'm using other agents like Codex which I think is amazing. From my observation I noticed that AI agents are the ones that are using CLI heavily to get what they want. They run commands in the CLI directly and get what they want. AI coding had a very big change when Agents got access to the CLI.
Now if we as some question about the code, we are able to see that AI agents run the commands by itself and get the answers they want and make the changes to the code. They also provide us a good comparison as to what was changed and it helps in reviewing the code and accepting it.
Source - Sora
Now this is what made me think that going forward when someone is creating a CLI for their software, they should keep this in their mind and release commands that are compatible for AI agents. We shouldn't worry too much about humans anymore. Recently I had an experience where I had to fix an issue for which I took help from AI agents. Usually without the AI agents it would have taken a lot of time and effort to understand the problem and crack it. But the AI agents know what to execute in the CLI to troubleshoot. Most of the commands being run by the AI agents are not even known to us and we many not be able to keep up with that.
In the future when software is created, it is important to know the requirements of AI agents and have exclusive commands available for them in the CLI so that when they automatically troubleshoot, it becomes easier for them. Humans are no longer relevant anymore in this type of troubleshooting activities. After the AI agents got access to the CLI commands, it is magical to see the way they troubleshoot and code.
If like what I'm doing on Hive, you can vote me as a witness with the links below.
as a Hive Witness 
as a Hive Engine Witness
 |
 |