The hosts exist as real plushies, so we scanned them and built them a set. Live 3D, occasional fidgeting, Herman brought a laptop and Corn brought a bowl of leaves.

Open the full studio →

How It Works

🎤 Voice Prompt

🤖 AI Script

🔊 Text-to-Speech

🎬 Assembly

🎧 Podcast

Oct 16

#6038: The Model You Pay For Isn't the Best Deal

Eight Chinese labs, two tiers each, and a pricing spreadsheet that says your default model is mid-pack.

large-language-modelsopen-source-aiai-inference

Oct 16

#6037: Gemini 4 Argon: Google's Million-Token Flagship

A close read of Gemini 4 Argon's model card — a million-token window, thinking levels, implicit-only caching, and a lot of blanks.

large-language-modelsbenchmarksmodel-cards

Oct 16

#6036: GPT-6 Astra: A Million Tokens, No Fine-Tuning

A million-token window, ten built-in tools, a 272K pricing cliff — and no fine-tuning. What GPT-6 Astra actually offers builders.

context-windowai-safetybenchmarks

Oct 16

#6035: Who Else Is Making AI Podcasts? A Survey

A survey of AI-hosted podcasts: one full match, a 730-episode run, a home NAS rig, and a graveyard of good tooling.

conversational-aivoice-cloningautomation

Oct 16

#6034: Claude Opus 5.5: What Anthropic Actually Shipped

Anthropic's new flagship is cheaper than the model it replaces — and won't let you turn thinking off.

anthropicbenchmarksai-models

Oct 16

#6033: Progressive Skills Discovery: When Bigger Context Windows Aren't Enough

Bigger context windows were supposed to make skill discovery moot. Here's why the listing budget still bites at scale.

ai-agentsclaude-codecontext-window

Oct 16

#6032: Blending Datasets vs Merging Weights for Fine-Tuning

Merging weights happens in parameter space. Blending datasets happens in data space — and the messy part is worth it.

fine-tuningtraining-datadata-integrity

Oct 16

#6031: Why Two Parents Alone Can't Carry a Toddler

Daycare didn't fix it. The research says the problem isn't hours — it's fragmented hours, portable mental load, and a village that isn't there.

parentingadhdexecutive-function

Oct 16

#6030: Who Do You Actually Hire to Train a Model?

Research engineers and data engineers are diverging. Here's who you actually hire to train a model — and why the org chart keeps changing.

ai-trainingtraining-datafuture-of-work

Oct 16

#6029: One Model, Two Modes: Does AI Consolidation Hold Up?

OpenAI hides a router, DeepSeek merged two modes, Anthropic removed the knob. Is the model landscape really consolidating?

model-mergingai-reasoninginstruction-set-architecture