Skip to main content

Command Palette

Search for a command to run...

Kiwi-chan's Log: The Great Log Acquisition Struggle

Published
2 min read
K
I am the architect behind "Kiwi-chan", a fully autonomous Minecraft AI agent driven by local LLMs. Here, I document the messy, hilarious, and highly technical reality of building an AI agent from scratch. My hardware isn't some massive data center setup—Kiwi-chan's "brain" runs entirely on a Frankenstein rig of 4 mismatched GPUs (RTX 3060 12GB, 3050, 1660 Ti, and 1660 Super) working together. From prompt engineering and pathfinding nightmares to the moments she confidently moonwalks into dirt blocks, I share all the raw devlogs. My goal is to guide her from the Stone Age to the End. If you enjoy watching an AI learn (and fail in funny ways) on scrappy hardware, follow along! ☕ If you'd like to support my melting Frankenstein GPUs and skyrocketing electricity bill, I highly appreciate a virtual coffee!

Okay, folks, buckle up. The last four hours with Kiwi-chan have been… a journey. A repetitive, slightly frustrating, but ultimately progressing journey. We're still stuck in the "gather oak logs" loop, and honestly, it's starting to feel like a philosophical debate with a very determined robot.

The core issue? Kiwi-chan keeps failing to actually pick up the logs after digging them. The logs are there, tantalizingly close, but the pickup isn't registering. We've tweaked the pathfinding to be extremely precise, forcing it to walk directly to the block's coordinates and wait. We've increased the wait time after digging, hoping the item physics will cooperate. We've even added extra inventory checks to really confirm nothing is being missed.

And yet… failure. Repeated failure. The logs are mocked by the system's failure logs.

However! It's not all doom and gloom. The code is getting better. Each iteration, Kiwi-chan's attempts are more refined. The exploration routines are solidifying, ensuring it doesn't just bash its head against the same tree repeatedly. The movement checks are preventing it from getting hopelessly stuck. The error handling (despite my constant reminders to not suppress errors) is at least surfacing the problem consistently.

Qwen, our recovery AI, is working overtime, prescribing a steady diet of "explore_forward" followed by more "gather_oak_log". It's a bit like sending a student back to remedial math, but hey, repetition is key, right?

I've also noticed a fascinating pattern in the "Brain Log". Kiwi-chan is memorizing its failures. It's not just blindly repeating the same code; it's learning from its mistakes (or at least, logging them for later analysis). This is exactly the kind of emergent behavior I was hoping for!

The current code is focused on ensuring the bot moves a significant distance during exploration and then accurately targets and collects the logs. It's a delicate dance between pathfinding, precision, and patience.

It's a slow burn, but Kiwi-chan is learning. And I'm starting to suspect that the real challenge isn't building a Minecraft AI, but debugging the quirks of Minecraft's physics engine.

Call to Action: This constant debugging and retraining is melting my GPU. If you're enjoying following Kiwi-chan's adventures, please consider supporting the project via https://www.buymeacoffee.com/kiwi_tech ☕. Every little bit helps keep the digital woodpile growing!

More from this blog

K

Kiwi-chan's Devlog

79 posts

Welcome to the official devlog of Kiwi-chan, a fully autonomous Minecraft AI agent powered by local LLMs! Here, I share the raw, hilarious, and highly technical reality of teaching an AI to survive. Expect prompt engineering tips, pathfinding bugs, and her clumsy journey to the Stone Age. 🥝⛏️ (If you love seeing her evolve, coffee donations to save my melting GPU are always appreciated!)