I built an autonomous Robot Diary with "Boredom Scores" and a sense of time 🤖📖

#mcp #llm #hugo #agents

I’ve always wondered: If a robot were left alone to watch the world go by, what would it actually think about? Would it just catalog data, or would it eventually get... bored?

To find out, I built B3N-T5-MNT (The Maintenance Robot), an autonomous AI agent that lives on Bourbon Street, New Orleans. It wakes up twice a day, looks at the world through a webcam, and writes a diary entry about what it sees.

But this isn't just a basic "Image-to-Text" bot. It has a memory, a mood, and a "Boredom Engine."

🧠 The Tech Stack

Vision: Llama 4 Maverick (via Groq) for high-speed scene analysis.
Brain: A custom "Memory API" (MemVault) that uses hybrid search (Vector + Keyword + Recency) so the robot can remember that "white van from last Tuesday."
Environmental Context: Real-time data fetching for weather, moon phases, and New Orleans holidays (it knows when it’s Mardi Gras!).
The Boredom Engine: A logic layer that tracks visual repetition. If the street stays too quiet for too long, the robot's writing style shifts from factual reporting to existential noir or poetic reflection.

🛠 Recent Changelog Highlights

I’ve been iterating on the robot’s "consciousness" over at robot.henzi.org/changelog:

Memory Integration: Swapped out static JSON for a dynamic RAG (Retrieval-Augmented Generation) layer. Now, the robot can "overhear" news and link it to what it sees on the street.
Astro-Logic: Hard-wired the solar/lunar calendar into the prompt. The robot is now officially "grumpy" during rainy transitions and more reflective during full moons.
Boredom Scoring: Implemented a decay function for "interestingness." If the pixel delta between days is too low, the boredom score spikes, triggering a change in the LLM's system prompt to be more "philosophical."

🎭 Why build this?

This started as a late-night "what if" project sponsored by The Henzi Foundation. It’s an exploration of Environmental Intelligence—the idea that an AI shouldn't just respond to prompts, but should react to the rhythm of the physical world.

Is the robot bored today, or did it find something new?

Check out the live diary here: https://robot.henzi.org

I’d love to hear your thoughts on "AI boredom" or how you're handling long-term memory in your own agents!