Latest Episode
Experimental Selected episodes are also translated into 8 other languages. Other languages →
Meet the Hosts
From the blog · Behind the scenes What One Episode Costs to Make About 27 cents with a premium model writing the script, about 16 cents with a cheaper one. Where the money goes, as of 31 October 2026, and how much a single sponsor could take off. 
#6617: Why CVAT Beats Generalists at Video Annotation
Born inside Intel in 2017, CVAT uses keyframe interpolation and track primitives to skip frames that generalist tools make you label by hand.
#6616: Small Labeling Tools That Outship Big Platforms
LabelImg is archived with 25,000 GitHub stars, while live single-machine tools like labelme and X-AnyLabeling ship AI updates weekly.
#6615: Quantization Quietly Deletes Fine-Tuned Skills
Three-bit GPTQ drops Llama-3.1-8B-Instruct by 14.5 points while the 70B version loses just 2, proving bit-width alone cannot predict damage.
#6614: How Humans Actually Grade Generated Video
While computer vision tools track objects across 100,000 frames, AI video evaluation still relies on two buttons and a five-second timer.
#6613: Where RLHF Preference Annotation Tools Stand
Open-source preference annotation specialist Argilla is in maintenance mode, leaving commercial platforms to ship the newest RLHF UIs.
#6612: How the Interface Shapes Data Annotation Tools
Audio and image annotation tools need heavy viewers, while LLM preference data collapses the UI to two text panes and a radio button.
#6611: Image Annotation When Experts Are the Bottleneck
The WikiChurches art historian annotated just 631 prototypical bounding boxes, narrowing 117 style classes down to four.
#6610: What Goes Into a Character-Voice LoRA Adapter
A character-voice LoRA adapter can be just twelve megabytes, yet it still requires a reference audio clip on every single generation.
#6609: Merging LoRA Voice Adapters vs Keeping Them Separate
Keeping fifty voice adapters separate takes 24 gigabytes of memory instead of 800 gigabytes for fifty merged models.
#6608: How Multi-Speaker TTS Models Keep Voices From Blending
Adding speaker conditioning to a vocoder can boost unseen-speaker quality by up to 95%, stopping the model from averaging every voice together.