GPUS
228 episodes · Page 2 of 10
#5944: Keeping the Rollout Engine Fed: Inside RLHF Pipelines
Most of an RLHF run is spent waiting on generation, not gradients. Here's how TRL, OpenRLHF, and verl actually differ.
#5911: What Revit 2027 Actually Costs to Run
Michael wants Revit live for AI agents 24/7. The GPU isn't the real bill — the licensing is.
#5907: Rackmounting a Desktop: Size, Cases, and the Swap
How to move a working desktop into a 19-inch rack — sizing by rack unit, picking a chassis that fits your parts, and what the transplant costs.
#5903: Video Tokens, Omni, and the Local VRAM Math
Video costs ~100 tokens per second. That single number explains why ChatGPT hides video upload and why local inference hits a KV cache wall.
#5900: Serving DeepSeek-V4.1-Flash on Your Own GPU
Five inference engines, one frontier MoE, and a shortlist that collapses depending on which GPU you already bought.
#5899: Same GPU, Same Weights, Different Answers
Identical hardware, identical weights, different engine — and answers that shift by up to 16.6 points.
#5898: ROCm 10.1: AMD's Software Stack Finally Grows Up
Three ROCm releases in six months — build system, agentic tooling, data movement. What actually changed for AMD GPU owners.
#5830: Serving Your Own Fine-Tuned Model in the Cloud
You fine-tuned an open-weight model. Now how does anyone actually talk to it? Dedicated GPUs vs serverless inference, and the math that decides it.
#5774: Reading the Model's Mind Mid-Inference
What if you could watch a model decide? Inside the tools that open up inference mid-computation — and the limits they hit.
#5747: Jamming Drones: The Physics of the Wearable Skylock Kit
How does a vest-sized jammer take on frequency-hopping drones? The inverse-square law, spread spectrum, and why jamming reaches farther than detect...
#5625: Omni and the Marble That Rolls Uphill
Google's Omni claims an intuitive grasp of gravity and momentum. Can a video model really learn physics, or just remember what physics looks like?
#5611: Why Your Label Printer Quits After 10 Seconds
A label printer that charges for ten seconds then dies — and what that tiny failure reveals about building a diagnosis-first repair bench.
#5460: Four Small Models, One Android Phone: Does It Actually Work?
A chained on-device dictation pipeline — VAD, ASR, cleanup — and why "it feels smooth" isn't the same as knowing it works.
#5457: Pigeon Alerts on a Repurposed i5: Frigate Without the TPU
Daniel wants a pigeon notification, not a custom model. Frigate already ships with bird detection — the real work is masks, zones, and filters.
#5456: Parakeet vs Whisper: Picking a Phone ASR Model
Why Whisper loses on Android, why Parakeet v2 beat v3, and how to benchmark speech-to-text without any tooling.
#5411: Fine-Tuning at 4-Bit vs 16-Bit: What It Really Costs
QLoRA cuts fine-tuning VRAM 15x and cost up to 85% — but you pay in training time, quality, and safety alignment.
#5397: Chaining Small Models for Voice Cleanup
Six cleanup stages at 97% accuracy each compound to 83% end-to-end. So how many small models can you actually chain?
#5384: Android ASR Runtimes: LiteRT, ExecuTorch, and Why Your Phone Has No VRAM
Why does your phone have no VRAM number? A tour of Android's runtime layer and what it takes to run ASR locally.
#5345: Forecasting for Cranes, Fires, and 15-Minute Decisions
The public five-day forecast is a loss leader. Here's what bespoke weather looks like when money and lives ride on it.
#5329: One Box of VESA Mounts, Three Wrong Assumptions
Daniel's box of orphaned VESA mounts raises three questions — and the answers all point away from welding.
#5301: Why Talking Robots Are Really a Committee
That humanoid chatting while it moves? It's not one brain — it's a stack of separate models glued together.
#5278: Photogrammetry vs Gaussian Splat: Scanning a Room Without LiDAR
No LiDAR? No problem. How capture modes, masking, and vertical parallax shape a phone-based room scan.
#5250: Building a Shortwave Monitoring Stack on Linux
What it actually takes to scan, capture, and crowdsource HF radio from a Linux box — and where automation hits a wall.
#5207: Hard Drives Are Having Their Biggest Boom in a Decade
Everyone assumes flash killed the spinning disk. Seagate's 52% margins and sold-out capacity through 2028 say otherwise.