Yet More Bangalore Sunsets
Hot (ish) on the heels of Even More Banglore Sunsets, here's the latest in my occasional series:
This time it was shot with the a7c and my favorite lens, the FE 20mm F1.8 G.
LHR CDG
San Francisco to Russian River
Hyperlapse of the drive from San Francisco to the Russian River and back again. Shot on Rivian Drive cam, editied in DaVinci Resolve and euro-trash music generated by Gemini.
Previously I've just done this with the front camera, but the Rivan has four and so I set up a multi-cam edit with 130GB of footage. I wish the side cameras were pointed a bit higher, but they are designed for parking rather than coastal views so I haven't opened a support ticket.
Seervo, an LLM Powered Robot
I just released Seervo, an open-source LLM-powered robot. The GitHub repo contains source code, a shopping list and 3D files to print the chassis.
Seervo sends an image from its camera to GPT 5.4. The LLM can decide to change the colors of four LEDs and to drive the motors. It has the objective of finding and entertaining humans while avoiding pets at all costs. The video below shows it mostly trying to escape from my dog:
The robot is based on an ESP32 microcontroller with a camera, some motors and a battery. The client code is MicroPython and it talks to an ASP.NET core web service that handles the LLM control calls. You could do everything on the ESP32 but it's easier to tweak prompts and see where it's going wrong with a local server. The server additionally stores memories so the robot can remember what it has been doing recently, and handles memory compaction so any really useful knowledge is retained in the context window.
The code was all written using Claude Opus 4.6. The chassis was designed in OpenSCAD using ChatGPT - something that has been a struggle before but GPT 5.4 can iterate on a 3D model with pretty vague directions.
Let me know if you build one!
Updated 2026-04-05 23:22:
Added an HC-SR04 ultrasonic distance sensor so the robot can now tell how much clear space there is in front of it. Tuned the instructions to use this data and also with formulae to convert distance and rotation to approximate motor run time. This all makes the robot a lot more confident in its movement.
Animation of US PM2.5 Air Pollution in 2025
This video shows hourly PM2.5 air pollution across the United States for 2025:
I did the same thing for 2023 using the Purple Air sensor network. Unfortunately they are now charging for access to the data, however the official AirNow network has an API and so I used that instead. The lack of sensors at the local level is frustrating, but it works fine for this nationwide view.
The 4th of July is pretty incredible (see the photo at the top of the post).
In San Francisco this was a pretty mild year for wildfire smoke. We did see a few days in mid-July from the Butler and Green fires. Canada was a big factor again, look particularly around June 3-5 and late July. And of course the Palisades fire, starting on Jan 7.
San Francisco to Lake Tahoe
Hyperlapse from San Francisco to Lake Tahoe:
I-80, SR 49, SR20, I-80 some more, SR 267 to Kings Beach.
This didn't look like a terrible idea the morning of December 26 (Boxing Day), the weather forecast was good. Google Maps predicted a little under four hours. Then it started to snow. And then there were many crashes, I-80 shut in both directions and the resulting jam took over six hours to clear.
Most of the hyperlapse runs at a 80-130x speedup. The jam portion is over 1000x.
In case you are insane here is the full drive, minus charging and food stops:
Shot on Rivian Drive Cam, post-processed as per this post, edited in DaVinci Resolve Studio, music hallucinated by Eleven Labs.









