I wanted a small, cheap computer that I could chuck into my bag for maintaining servers and writing emails. So I modded my ancient eee PC from 15 years ago: it has an SSD drive and a USB-C charging port. Works plenty fast enough for that! #retrocomputing #raspberrypi #modding #upcycling

It surprised me that there weren’t any free apps for just putting on some nature sounds to fall asleep to. So I bought a professional recorder 2nd-hand and made my own app. Now it’s a hobby and a good excuse to visit wild areas. #sound #soundscapes #chill

Early experiments with DeepSeek-R1 for my robot’s take planning. Takes slightly longer, but benchmark scores are way better than Llama3.2. My goal is to replace as much human-coded logic as possible with a vanilla locally-hosted LLM. #ai #llm #deepseek #foss

Love flexible TPU for 3d printing, can’t say I’m a fan of the form factor of my old smartwatch prototype any more… think I’ll wait for the new Pebble #smartwatch #tpu #3dprint

“What can I do with these?” [holds up scissors] More vision testing with llama3.2. I’m going to cut out a lot of “hand coded” logic, and replace it with this multimodal LLM + prompts. Video cuts out about 15 sec of waiting for the LLM. #ai #llm #robotics #robot #esp32 #arduino #iot #hardware

LLM-powered robot, remembering the past and inferring ambiguous commands to manipulate objects. Was using a slow GPU at the time, so inference pauses are edited out. #llm #ai #robotics #opensource #robot #programming #iot

Time lapse of building my robot motherboard v1.0 and the first version of the robot with only 4 legs #ai #robotics #robot #hardware #esp32

My robots having a chat. Haven’t got the movement working again for both at the same time, yet. Running local llama3 8B instruct, source and schematics on my github #robotics #opensource #ai #llm #robot

I wish I had this in the 80s. ChatGPT on a Casio handheld computer. #chatgpt #retrocomputing #casio #retro