Calling out some bad behaviour by @_rexliu from @gen_analysis on LinkedIn - amateurish AI Image Edit to steal someone else's post and claim it for themselves Two days ago, friends at @exolabs released an awesome post with @nvidia + @Apple and using both hardware to speed up LLMs on-device. And then, I saw a post on LI by @_rexliu, claiming they "just fine-tuned gpt-oss-120B (QLoRA 4-bit) on a DGX Spark..." (see screenshots). Super familiar setup, strikingly similar table, strikingly similar plant in the background, remarkably similar cable placements. And no device ID from NVIDIA below the sticker. (amateur mistake) Yeah.... These guys definitely saw @alexocheema's post on LI going viral, stole the post image, and AI-edited it, so they can steal some unearned praise for running these cool innovations. (And no credits to @exolabs) These types of AI image edits to steal credit will become increasingly easy and common now. Be careful out there.

Calling out some bad behaviour by @_rexliu from @gen_analysis on LinkedIn - amateurish AI Image Edit to steal someone else's post and claim it for themselves Two days ago, friends at @exolabs released an awesome post with @nvidia + @Apple and using both hardware to speed up LLMs on-device. And then, I saw a post on LI by @_rexliu, claiming they "just fine-tuned gpt-oss-120B (QLoRA 4-bit) on a DGX Spark..." (see screenshots). Super familiar setup, strikingly similar table, strikingly similar plant in the background, remarkably similar cable placements. And no device ID from NVIDIA below the sticker. (amateur mistake) Yeah.... These guys definitely saw @alexocheema's post on LI going viral, stole the post image, and AI-edited it, so they can steal some unearned praise for running these cool innovations. (And no credits to @exolabs) These types of AI image edits to steal credit will become increasingly easy and common now. Be careful out there. @garrytan, you have some bad behaviour in your YC cohorts.

Clustering NVIDIA DGX Spark + M3 Ultra Mac Studio for 4x faster LLM inference. DGX Spark: 128GB @ 273GB/s, 100 TFLOPS (fp16), $3,999 M3 Ultra: 256GB @ 819GB/s, 26 TFLOPS (fp16), $5,599 The DGX Spark has 3x less memory bandwidth than the M3 Ultra but 4x more FLOPS. By running compute-bound prefill on the DGX Spark, memory-bound decode on the M3 Ultra, and streaming the KV cache over 10GbE, we are able to get the best of both hardware with massive speedups. Short explanation in this thread & link to full blog post below.