🌱 “If you aren't growing into the future, you are shrinking into the past.” - David Droga

Hello Reader,

Have you ever heard of the Diffusion of Innovation? It's a bell-curve distribution of how technology is adopted over time.

OpenClaw is not cutting-edge technology. It is bleeding-edge technology. Using it is bloody and difficult.

This is what the founder of OpenClaw, Peter Steinberger, published on his blog this week:

When I had both of my OpenClaw agents running on 4.29, they were basically unusable. I was spending most of my time troubleshooting, which, I must admit: I'm not good at, and I don't enjoy.

So I let the critters sleep for a few days.

I knew the problems were well-known; X and Reddit complaints were plentiful. I also knew I would eventually have a simple update that fixed most of these issues.

This left me with two options:

1. Keep trying to make glitchy software work, right away, or
2. Wait for patches and updates that made my frustration obsolete.

There is plenty of uniquely-human work I could be doing, instead of wrestling with software with known bugs. So I ignored my OpenClaw agents, and went to work on those uniquely-human tasks:

• Call other humans on the phone
• Decide on my next launch strategy
• Catch up on the laundry and chores around the house
• Deliver the slides and replay from a recent Christchurch AI Meetup with my thoughts (and AI-augmented images) on a LinkedIn post​
• Develop the IP for a new AI Playbook full of AI games

I almost took this downtime to go and try the Hermes agent. But I'm glad I took a break from agent management, because I had a backlog of human management to handle.

OpenClaw 5.2 fixed a lot of the issues I had (Discord, gateways, and plugin config, mostly) and I picked up on my OpenClaw projects where I left off.

I even left my Mac Mini accidentally unplugged overnight. It was a refreshing step back into what life was like a few months ago, before I had these overnight agentic capabilities.

💡 The Nightly Build

Every night, before I go to sleep, right before I turn my phone on Airplane Mode and plug it in to charge, I open up Telegram and ask each of my OpenClaw Agents, in turn:

I yawn, I stretch, I give it a minute. I don't read the plan. (I just have to make sure /think is at least 'medium' and /reasoning is 'on')

Then I respond with:

Each OpenClaw agent I run has a specific Mission Statement, and has the authority to autonomously pursue the achievement of its goals. I don't need to know all the steps it's taking, but I do need to prod it to keep it busy now and again.

It's like spinning plates. Right now, OpenClaw agents need constant attention. The promise of a 24/7 AI Agent controlling its own computer is that it accomplishes work without you, but it often needs multiple daily sessions with you, its Claw Manager.

The Nightly Build ensures my OpenClaw agents will continue working while I sleep. So even if I don't spin the plate myself, they know to keep moving.

My agents also use the Self Improving Agent skill to review their performance and their problems every morning at 2am, write a report with recommendations, and update their MEMORY.md files with fixes.

🤖 The OpenClaw Owners Manual

I've been providing custom OpenClaw installations in New Zealand, to help businesses hire their first AI agent. It's been a lot of fun, and I've been hosting weekly Open Office Hours for my OpenClaw clients to help them troubleshoot the difficulties of keeping their agent online.

Throughout this process, I have collected a variety of basic tasks that every OpenClaw owner needs to be proficient in managing their AI Agent.

I'm building this in the AI Coaching Academy, and I'll need a few beta testers before I launch it.

If you're running your own OpenClaw installation, reply with a lobster emoji 🦞 and I will get you free access before I launch it.

📰 New AI News This Week

• In a study comparing AI and ER doctors, OpenAI's o1-preview gave more accurate diagnoses than humans
• OpenAI released Codex Pets, animated critters that represent your agentic workflows passively in the background
• A Chinese court has ruled that replacing a worker with AI is not legal grounds for termination

👓 What I’m Reading

• The AI Blueprint for Aotearoa (May 2026 Edition)
• Paul Millerd's Strategy Consulting Skills​
• My runsheet for the all-day EPIC AI Conference I am hosting for TechWeekNZ in Christchurch on Thursday, 21 May

Upcoming Events in the AI Coaching Academy

Free events I am hosting as part of TechWeekNZ:

Monday, May 18 at 6:30 PM GMT+12 Managing AI Agents: The Next High-Value Skill Set AI agents are changing the way work gets done. As agentic AI systems become more capable, our real advantage will not come from technical proficiency. Our advantage will come from knowing how to manage AI agents well. The ability to scope agentic projects, manage autonomous systems, evaluate their performance, and design the information... RSVP

Wednesday, May 20 at 12:00 PM GMT+12 AI Governance & The Future of Work in New Zealand AI is already in the room. The question is: who’s in charge? Across Aotearoa, organisations are moving fast—from experimentation to real-world AI deployment. Productivity gains are clear. Investment is accelerating. But governance hasn’t kept up. Most leaders are now responsible for AI outcomes without clear frameworks, shared language, or... RSVP

Friday, May 22 at 12:00 PM GMT+12 AI Coaching Power Hour A facilitated interactive experience to think better, together — with the help of AI. What happens when you blend human connection, coaching frameworks, and an AI that listens without judgment or ego? You get a power hour that jump-starts clarity, confidence, and momentum. Join Caelan Huntress for an AI Coaching Power Hour. This session is... RSVP

Keep on growing!

- Caelan Huntress ​[email protected]​

​​​​​