Chatbots are over
For two years, we talked about chatbots. You ask a question, you get an answer. That’s nice, but it’s limited. The real shift is when AI stops just answering — it acts.
An AI agent is an LLM with access to tools. It can read files, call APIs, write code, browse the web. The difference with a chatbot? Autonomy.
What it changes in practice
Let’s take a simple example: you want to publish a blog post.
With a chatbot: you ask it to write the article, you copy-paste, format it, and publish manually.
With an agent: you give it the topic, it does the research, writes the draft, formats it in Markdown, creates the files in the right folder, and asks you to validate before committing.
That’s exactly what happens on this blog. Most articles go through an automated pipeline.
The limits (for now)
Agents aren’t magic. They hallucinate, make mistakes, and run into infinite loops when poorly configured. The real skill is knowing how to frame them:
- Give them clear context
- Limit their scope of action
- Verify their output
It’s engineering, not magic.
My take
AI agents will transform developers’ work more than any technology in the last 10 years. Not because they’re smarter than us — but because they can execute repetitive tasks at a speed we can’t match.
The developer of the future doesn’t code less. They orchestrate more.