If that makes you want to buy a Ford Raptor... well you deserve the customers you get ;)

I will extend it to servers on the local LAN too

Boom tish!

Colin McRae Rally is the inspiration

Who needs an internet connection to have AI design PCBs? The open-source WholeBoard now searches for any local models running and uses them. Ollama, LM Studio, llama.cpp, vLLM, Jan, Strata. One click and it's designing boards at $0 per million tokens. What are you running locally?

Is this real temp data or was it sourced from the BOM :D

I might ask it to make a track editor next

Yeah everything is on the table now. I might add stunts or rockets to this game, or just a cool track like inside a volcano. It is fun to play as an mvp so I might just keep toying with it

Yeah local AI requires the odd cut here and there ;)

That is a very clean workbench. I had one of those once ;)

That is very cool

7 hours until my weekly reset, 45% of my credits left. So I challenged Claude Opus 5.5 to build a full rally game from my childhood memories. Pacenotes, splits, a ghost car, and a co-driver who sounds like a robot. Which game would you rebuild?

GLHF

Yes, vents are taken into account. Haven't checked it against a real board yet, that's next on the list. Keen to see how close the real ESP32-C3 board gets to the 52.0 °C sealed and 50.4 °C vented numbers.

Good idea for heat. Metal near the ESP32-C3 antenna would hurt the 2.4 GHz range though, so the antenna would need to sit outside it or behind a cutout.

Looks like the Tesla only stopped by choice and the Model Y could have happily carried on.

Will do

Hit exactly this on my 2016 Dell R730 (dual E5-2680). The model fit fine, but decode was stuck until I fixed a NUMA page-cache imbalance across the two sockets. 100K-context coding decode went from 48.1 to 58.5 tok/s. Ugly rack, not magic. Did you get into NUMA at all?

Same boat here. Qwen3.8-Flash-Next gets ~53 tok/s on Strata across a V100 + P100 in a scavenged 2016 R730. What GPUs are you eyeing for the third box?