Just tested the new NVIDIA DGX Spark for the first time today and it's a great device! 🔥 And it's small! I immediately ran a real head-to-head against our RTX 5090 on the exact same workload. - The RTX 5090 is roughly 6× faster in raw tokens/sec - The DGX Spark still delivered a very usable 35 tokens per second on Qwen Coder Next 80B At first the speed gap felt noticeable... but then you realize what's actually happening. The 5090 is an absolute speed demon when the model fits comfortably in 32 GB VRAM. The DGX Spark, however, brings 128 GB of unified memory. That single difference changes everything. I could load the full 80B model with room to breathe: longer context, lighter quantization without constant memory gymnastics, no compromises. This isn't about winning the pure performance crown. It's about making frontier-scale local AI actually practical on your desk. And it's quiet, private, no cloud bills, no latency and no data leaving the building. The DGX Spark isn't trying to replace high-end GPUs. It's redefining what "usable" means for serious developers and researchers. Already planning my next experiments. Can't wait to push it further. Anyone else got their hands on the DGX Spark yet? #DGXSpark #NVIDIA #LocalAI #CodingAgents

We are amazed to see another paper 📜published featuring #FireSOM for the prediction of behavioral categories from 🦇multi-axis acceleration data. Congratulations to the authors! conbio.onlinelibrary.wiley.c… #Firetail14 #sbgResearch