← Back to Main Site

Project Journal

Notes, experiments, and documentation

Setting Up ClawdBot (Moltbot) on Raspberry Pi 3

Today I started setting up ClawdBot (also known as Moltbot) on an older Raspberry Pi 3. This is a project I've been meaning to document for a while.

Initial Setup

Details about the initial setup process will go here...

AI API Costs

I've been working with the Claude API and discovered it gets quite costly with anything above Haiku. I looked into Gemini as a potential alternative, and I'd likely try that, but I ended up going in two directions at the same time:

  • Venice AI $VVV token - Staking their coin leads to daily credits for AI usage
  • Claude Pro subscription - Leveraging a subscription instead of raw API calls

Claude Code Subscription Idea

Using a Claude Code subscription to avoid some of the API costs. This raises an interesting question: if I can run ClawdBot and give it knowledge of Claude Code, can Claude Code run the raptor personality to leverage Anthropic models at the subscription level instead of churning API tokens directly? Will that save money?

Challenges

Notes about any challenges encountered...

Next Steps

  • Item one
  • Item two
  • Item three

More details to come as the project progresses...

The AI Bug Bounty Gold Rush — Are We Chasing the Wrong Bugs?

Traditional vulnerabilities still account for 78% of successful exploitations, yet bounty payouts for AI bugs have jumped 340%. Are researchers chasing money instead of risk?

Read full article →

Building an SAO for DEF CON — 555 Timer LED Chaser

A purely analog Simple Add-On for DEF CON — a 555 timer driving a CD4017 decade counter that chases eight LEDs across a custom 60mm × 60mm PCB. No microcontroller, no firmware. Just power and ground.

Read full article →

Prompt Engineering & Vulnerabilities

A deep dive into prompt injection techniques, jailbreaks, and vulnerabilities — from basic instruction overrides to recursive multi-model exploits. Covers 12 attack patterns including role-play jailbreaks, context overflow, cognitive hacking, and the defined dictionary attack.

Read full article →