welcome
Welcome to my blog.
I’ve been stuck in a haze of indecision for the past few months. What should I build? I use AI heavily 8 hours a day, 5 days a week at work. Should I resist using it in my side projects, and write everything by hand the old-fashioned way?
Then comes along Clawdbot. I mean Moltbot. I mean Openclaw (thanks Anthropic). Built predominantly by one bloke and his army of AI agents (what is the collective noun for a group of agents??). I immediately wanted to know more about the human behind this story. How does he think about building in 2025/6? How much does he hand off to agents? Is he as obsessed with multi-agent complexity as some of the more evangelistic AI preachers?
It was refreshing to see that he keeps things relatively simple. Yes he has 9 agents (i.e. 9 harnesses) open at a time, but he pushes straight to main, keeps simple rules in an AGENTS.md file, doesn’t believe in git worktrees and focuses on short, iterative feedback loops. This is a philosophy I can get behind. Keep it simple stupid. Overcomplicating the tooling results in less cognitive load available for what matters; system design, architecture, UX and taste.
Anyways, what I’m taking away from Peter is that there is no right way to build. There is no right time to build. But the only way to find passion, learning and fulfilment as a builder is to start now. Chip away at it. Make mistakes.
With that being said, I am going to build projects in a myriad of ways. Some will be fully hand written, where I challenge myself to build fundamental servers, protocols and apps. Others will be deployed with me barely having looked at the source code. It really depends on my objective behind creating each piece of software.
~lachie