PinnedInData Science CollectivebySebastian Buzdugan·Oct 1Your Voice Agent Transcribes the Wrong Person: I Measured ItI played two people talking at once into Whisper 80 times. In 29 of them, the user’s sentence vanished and the transcript belonged entirely…A response icon12A response icon12
PinnedInAI, ActuallybySebastian Buzdugan·Mar 28Why a $500 GPU Can Beat Claude Sonnet on Coding BenchmarksA response icon13A response icon13
PinnedInAI, ActuallybySebastian Buzdugan·Jan 7A 30B Qwen model runs in real time on a Raspberry Pi, here’s why that mattersClick here to read this article for free…A response icon16A response icon16
PinnedInAI, ActuallybySebastian Buzdugan·Mar 8, 2025What Are MCP Servers? The New AI Trend Explained for EveryoneArtificial intelligence is taking a big leap beyond just answering questions — it’s now reaching out into the world of our data and tools…A response icon2A response icon2
Sebastian Buzdugan·6h agoI Reproduced the Blackwell GPU Fingerprint on a Laptop in 14 SecondsA report published on November 12 claims six Chinese frontier models were initialized on export-controlled NVIDIA Blackwell hardware, and…A response icon2A response icon2
Sebastian Buzdugan·9h agoI Benchmarked CloudMounter Against rclone on 6.5 GB of S3 DataOne app opened a 5 GB model checkpoint from S3 in 3.1 seconds. The same app needed almost fifty minutes to read 3,515 small images from the…A response icon1A response icon1
Sebastian Buzdugan·3d agoOne Sentence Took Claude’s Refactor Rate From Under 10% to HalfAnthropic measured a single line of system prompt that moves Claude Fable 5.2 from merging duplicate code in under one try in ten to…A response icon4A response icon4
InData Science CollectivebySebastian Buzdugan·3d ago100 parallel Harbor rollouts on Tensorlake: 47 rollouts/minute at 37¢ per batchFrom 10 to 100 concurrent sandboxes: 7.9x the throughput, 300 of 300 passed.A response icon1A response icon1
Sebastian Buzdugan·4d agoForwardAgent Defaults to No. The Interesting Bug Is What Happens When You Say Yes.Click here to read this article for freeA response icon1A response icon1
Sebastian Buzdugan·6d agoUnplugging the Laptop Made Inference Four Times Faster.On an RTX Spark laptop running native Linux, GPT-OSS-120B generates about 15 tokens per second plugged into the wall. Pull the power cable…A response icon1A response icon1
