Press Release

Nov 23, 2025

We are entering a golden age of computing. With modern Large Language Models (LLMs) capable of generating robust, functioning code, a new class of software applications becomes possible—applications developed by agents, not just humans.

Software engineering is on the cusp of a major shift: from developers writing applications directly, to developers building, evaluating, monitoring and guiding agents that write maintainable applications on their behalf. This new discipline is what I call Agent Engineering.

An essential part of using a coding agent effectively is being able to trust it—and trust comes from understanding. When you know how an agent works, you understand why it succeeds and why it fails, which makes it far more powerful and predictable. Which is why I feel the Agent Framework code should not be opaque.

So if you’ve ever been amazed by coding agents like Claude Code or Codex—how they craft complex software from simple prompts—and wondered how they actually work, or even wanted to build your own, this series is for you.

I will be live-streaming the creation of an Agent Framework from scratch in Go. We’ll start with frontier models, then progressively transition to open-weight, locally hosted models. The end goal: to empower anyone—including non-engineers—to build complex, maintainable software through agentic workflows rather than one-off “vibe coding.”

---

FAQ

What is an Agent Framework?

An Agent Framework is the set of libraries, services, evaluation tools, and runtime components that make it possible to build, observe, and improve agents.

An Agent is an instance of a process that perceives its environment and takes actions—often through tools or APIs—to achieve a goal.

---

Where and when are the streams?

Every Sunday 9pm-10pm Singapore Time (SGT)
On Youtube. Coming soon on Twitch.

What the tech-stack?

Golang as the programming language for Agent Framework development.

What are the prerequisites?

Basic familiarity with Go programming language will be beneficial to following along.
But it’s not a prerequisite. I’ll try my best to translate the Go concepts in terms of other languages like Java, Python, Javascript etc.
If you are completely new to programming, I think following along with the stream will be a good learning experience.
No deep ML knowledge is required.

What will the first and second episodes cover?

Episode 1

• Setup the project
• Write the core Agent loop
• Integrate with Claude
• Write simple tools to read file.
• Test: Can the agent explore a project repository and summarize?

Episode 2

• Support ChatGPT, gemini models
• Enhance tool set to search, write files.
• Test: Can the agent edit a correct file to

---

How will we validate that the coding agent works?

You can’t build a effective Framework unless you have a good usecase
So as a dogfooding exercise, We will be build Chess Learning app using our coding agent.
The goal of the Chess Learning app is to improve my chess knowledge to help me to improve my chess rating from 1200 to 1500.

Will we use other coding agents to help develop the framework?

No. We will not use something like Claude Code or Codex to build Agent Framework.
When our coding agent becomes mature enough we will use it to develop itself. Until then, we will develop a toy-application.

What about other open source Agent Frameworks. Why not just use those?

True. There are gemini-cli, Google’s Agent Development Kit (ADK), Microsoft’s Agent Framework etc. which are open source
But what I’m trying to do with the agentengineering.dev project is discover the new engineering practices with agents, and I believe that aim would be better served by developing the Framework from scratch.