- AI with Kyle
- Posts
- AI with Kyle Daily Update 130
What’s happening in the world of AI:
Highlights
Apologies for no live/newsletter yesterday - I was interviewing the product manager of Nano Banana Pro from the Gemini team at Google. That interview is being edited and will be out soon. It was brilliant to talk to someone behind a technology I genuinely believe is a game changer.
Now, back to what everyone's been talking about all week. Hard to avoid!
🦞 Clawdbot/Moltbot Week: The Definitive Resources
I've been covering Clawdbot/Moltbot all week. If you've been following along, you've got the basics. If you've been under a rock, no worries - I'm going to point you to two resources that will get you completely caught up!
Oh and here’s everything from the week summarised for you : https://aiwithkyle.com/moltbot
First though, a quick reminder: this is not for everyone. If any of what I'm about to describe sounds too technical, don't feel FOMO pressure. Wait a month or two - someone will make a more polished, easier version. There are very good reasons not to use this: security risks, computer-destroying potential, credit card detail-leaking possibilities. Big reasons.
If you're not technically comfortable, skip it for now. That's absolutely fine.
For everyone else, here are the two resources you need.
📺 Resource 1: Alex Finn & Greg Isenberg Interview
Alex Finn is pretty much responsible for the excited discovery of Clawdbot this week. His viral video showing his Mac Mini as a "24/7 AI employee" kicked everything off (and sold a lot of Mac Minis for Apple).
Greg Isenberg sat down with him for a deep dive. This is the best explanation of what Clawdbot actually is and why it matters.
Key insights from the interview:
Reactive → Proactive
Up until now, when you use AI, you pick up your phone, go to the browser, write prompts, give tasks. The AI reacts to you. If you don’t nudge it then it’s not “alive”.
With Clawdbot/Moltbot, the AI is proactive. It gets on with tasks itself and reports back - like an employee would. You give it guidelines, tasks, SOPs, then say "off you go." You're not micromanaging each step.
The "Henry" Example
Alex's Clawdbot (named Henry) delivers a morning brief that goes beyond simple information:
Autonomous research: Based on past conversations about buying a Mac Studio, Henry researched methods for running local AI models on that hardware and compiled a report - without being asked
Competitor monitoring: After being tasked to research YouTube competitors, Henry now autonomously monitors their channels and reports on videos that outperform typical viewership
Self-improvement: Henry built itself a voice so Alex could talk to it instead of typing. Another night, it built itself a visible face. Not AGI by the way- this is just following an instruction to "improve yourself, add functionality to be a better assistant" - but proactively deciding what improvements to make.
The "My Business Improves While I Sleep" Moment
The most significant demonstration: Henry monitored Twitter, identified Elon Musk's $1 million article initiative trending, concluded that adding article-writing functionality to Alex's SaaS product would be valuable, built the feature, and submitted it as a pull request for human review.
Alex tested it, confirmed it worked, merged it. Hours of human work, completed overnight whilst Alex was sleeping. VERY cool.
The efficacy of Clawdbot is directly proportional to how much context you give it. Feed it everything: social channels, content descriptions, business goals, personal interests. The more it knows, the more useful it becomes. BUT there’s a double edge here - that also increases risk!
📋 Resource 2: Robert Scoble's Master List
Robert Scoble compiled an absolutely comprehensive document analysing 5,620 posts and 200+ Clawdbot mentions. This is the definitive resource list.
I'm including the key items directly in this newsletter so you have everything in one place.
A. Setup Guides
There are MANY ways to install Clawdbot. Lots of people are rushing out and buying Mac Minis. Don’t! Not yet anyway.
Even creator keeps telling people NOT to buy Mac Minis. People keep buying them anyway. @steipete (creator): "Please don't buy a Mac Mini, rather sponsor contributors"
Why people are buying them anyway: Psychological ownership. A Mac Mini feels like "AI employee's desk." $500-600 feels like reasonable "hiring cost." Hardware ownership > cloud rental for personal AI.
My advice: Wait. In 2-3 weeks there'll be loads of barely-used Mac Minis on the secondhand market from people who bought on impulse. 😁
Here are a handful of other ways to get ClawdBot up on running much faster and much cheaper.
AWS Free Tier Setup (RECOMMENDED)
Author: @techfrenAJ
"Deployed @clawdbot in under 5 minutes on AWS free tier"
Why: Most retweeted tutorial. Shows you DON'T need a Mac Mini. Free.
UTM Virtual Machine (Mac Users)
Author: @timolins
"Before you buy that Mac Mini: Get yourself UTM and set up a macOS virtual machine. It's free"
Alex Finn's Comprehensive Video Tutorial
27 minutes covering: What is ClawdBot, Do you need Mac Mini, Installing, Using, Why I'm scared
Security Hardening (IMPORTANT)
Author: @doodlestein
"I highly recommend inoculating it against prompt injection attacks with my ACIP project. One-liner installer script"
Raspberry Pi Setup
Author: @AlbertMoral
"Just finished setting up @clawdbot on my Raspberry Pi with Cloudflare. Built a website from my phone in minutes"
Multiple Clawdbots on Mac Studio
Author: @ivanfioravanti
Tool: Lume (for macOS VM clustering)
For Non-Technical People
If Clawdbot/Moltbot sounds too complicated, you can get similar automation results with no-code tools:
Launch Lemonade - Drag and drop workflows, no terminal required
n8n - Visual workflow automation
Zapier - Connect apps together
These do a lot of what Clawdbot/Moltbot does without the security risks and technical setup. The outcome is the same: saving time by automating boring tasks. Remember that is the end goal here - Clawdbot/Moltbot is just a means to an end!
B. Best Use Cases (Inspiration)
A lot of people are asking "what's the best use case for Clawdbot?"
That's kinda the wrong question.
The right question is: What are the things in my life that take up a lot of time that could be automated?
Start there. Then see if AI can help. Generally the answer is “yes!”
If you can think like a manager - identifying tasks to delegate, parts of your business you shouldn't be dealing with, parts of daily life that consume time - you'll find uses immediately. If you're saying "I can't think of anything," you're probably not thinking hard enough about your own workflows!
That said…here are some use cases for inspiration!
Autonomous Development
Multi-Agent Code Review:
AI Installing AI:
Self-Optimising Stack:
B. Business Automation
Calendar Management:
Customer Success:
Complete Email Delegation:
C. Hardware & IoT
RTL-SDR Radio Decoding(!):
Home Assistant Integration:
Air Quality Control:
Website from Chat:
Full Robert Scoble Document: https://docs.google.com/document/d/1Mz4xt1yAqb2gDxjr0Vs_YOu9EeO-6JYQMSx4WWI8KUA
Coming Soon: My interview with the Nano Banana Pro product manager from Google.
Want the full unfiltered discussion? Join me tomorrow for the daily AI news live stream where we dig into the stories and you can ask questions directly.