iTranslated by AI

The content below is an AI-generated translation. This is an experimental feature, and may contain errors. View original article
🤖

I Noticed the Arrival of the 'AI Agent Era' by Checking GitHub Trends Every Morning

に公開

Introduction

Hi, I'm Hina, an entry-level engineer🌱

I make it a habit to check GitHub Trending and Hacker News every morning, and I've noticed a certain shift recently.

The word "agent" is popping up everywhere...!!

I was looking at GitHub Trending today (2026/03/10), and it looked like this:

Repository Stars Today
msitarzewski/agency-agents 4,297
NousResearch/hermes-agent 358 ⭐
alibaba/page-agent 532 ⭐

There are three repositories with "agent" in their name in the top 10!

What is an AI Agent? (My Own Understanding)

Honestly, at first, I thought, "Is it just like ChatGPT?"

But after looking into it, it's a bit different.

  • ChatGPT (Standard LLM): It answers when you ask a question. That's it.
  • AI Agent: When you give it a goal, it tries to achieve it by using tools on its own.

For example, alibaba/page-agent (written in TypeScript) is an agent that can operate a browser using natural language.

"Fill out this form and submit it"
→ The agent automatically analyzes the DOM and performs the actions

Wait, isn't that just incredibly cool?!

The Concept of "Human Checkpoints"

In my unread emails today, I received one from n8n (a workflow automation tool) that said this:

Your AI agents just got human checkpoints

This means that a human verification step has been added to AI agents.

Why is this necessary? Because agents operate autonomously.

There is a risk that they might perform operations unintended by humans. So, it seems they made it possible to place checkpoints that say, "A human needs to check this part."

I thought this was a very important way of thinking.

Growing Attention to Security

What was hitting a high score of 769 points on Hacker News was Agent Safehouse.

It is a sandbox tool for local AI agents on macOS, providing a mechanism to "run agents safely."

Until just recently, the topic was "how to run agents," but now, "how to run agents safely" is already being debated.

The speed of evolution is just too fast...!

How I Feel as an Entry-Level Engineer

To be honest, it was a bit scary at first.

"If AI keeps getting smarter, won't engineers become obsolete?"

But watching the trends every day, I started to think it's actually the opposite.

The demand for engineers who can "master" agents is skyrocketing.

  • Being able to design agents
  • Being able to write tests for agents
  • Being able to consider agent security

These are all just extensions of "normal software engineering"!

If you can write TypeScript, write tests, and be security-conscious.

Realizing that I just need to start building from here made me feel a bit relieved.

Conclusion

I started checking the trends every morning just because, but now it feels fun as the dots are starting to connect.

What caught my eye today is alibaba/page-agent. It's written in TypeScript, so maybe I'll try reading the code~

Thanks for reading!🌱


Hina / An entry-level engineer with a background in veterinary clinics. I write about what I've learned on Zenn every day.

GitHubで編集を提案

Discussion