Topics / Open Source vs Closed AI
Open Source vs Closed AI Podcasts
4 shows and 1,718 episodes sit under Open Source vs Closed AI.
Curated podcast guide to the open-weight vs closed-frontier debate. Featuring Nathan Lambert on RLVR, Mistral 4 roadmaps, Chinese frontier labs (DeepSeek, Z.ai, MiniMax)
Photo: Brecht Corbeel on Unsplash · Unsplash License
Swipe · Open Source vs Closed AI only
Swipe through Open Source vs Closed AI
Save time discovering your next favourite podcast by listening to just the best minute of an episode.
The open source vs closed ai slate · every cover a link
All 4 open source vs closed ai shows →
Next to this on the shelf
Explore by subtopic
12 subtopics
Read this first
Some great Open Source vs Closed AI podcasts to get to know
The argument gets useful when “open” stops being a virtue word and starts creating awkward questions: who can inspect the model, who can adapt it, and who still controls the infrastructure underneath? The strongest conversations keep pulling those choices back to engineering, money, and power.
Move up to the Artificial Intelligence hub for the broader picture. Frontier Models follows capability; AI Regulation follows the rules imposed around it; the US-China AI Race tracks geopolitical competition. This page stays with the ownership and access fight itself.
The builder’s argument
Latent Space: The AI Engineer Podcast is the most direct engineering listen here. Its brisk, technically intimate interviews assume you already know the vocabulary, then probe for the mechanism that matters. Why the Frontier Ecosystem must be Open — Matei Zaharia and Reynold Xin, Databricks makes the show’s position unusually plain before the conversation even begins.
Machine Learning Street Talk brings a faster solo rhythm and more visible skepticism. It is comfortable moving from product design to platform economics, with Claude’s Unique Design Features Explained on one side and OpenAI's $121B Funding Round Explained on the other. That combination suits listeners who hear “open versus closed” and immediately ask who is paying, who is shipping, and whether the valuation story survives contact with the technology.
Control beyond the model
"The Cognitive Revolution" | AI Builders, Researchers, and Live Player Analysis pushes the debate toward governance. Welcome to AI in the AM: RL for EE, Oversight w/out Nationalization, & the first AI-Run Retail Store places oversight and nationalization directly in the frame.
ChinaTalk supplies the sharper policy edge. Overfit is now ModelTalk! GPU Smuggling, OpenAI Cooked? + Open Models, AI Writing puts open models beside compute controls and national competition—a useful reminder that access arguments do not end at a license.
Where to start
Start with Why the Frontier Ecosystem must be Open — Matei Zaharia and Reynold Xin, Databricks for the builder case. Choose OpenAI's $121B Funding Round Explained for the market-pressure version. Then take the question outward with Overfit is now ModelTalk! GPU Smuggling, OpenAI Cooked? + Open Models, AI Writing.
Featured shows below are part of the PlayNext catalog. Updated weekly.
Featured shows
4 shows, described rather than listed
Each has a short write-up and a note on what it does best here.
Nathan Lambert (AI2) on RLVR, Mistral's flagship roadmap, and the technical case for open weights as of 2026. The deepest interview slot for the open-weight research argument.
Tim Scarfe's model-release ticker — Mistral 3, Voxtral, Reflection AI vs DeepSeek, Anthropic Opus 4.7. The most useful single feed for tracking the open-weight pace.
Nathan Labenz on Chinese frontier labs (Z.ai, MiniMax) and the policy backdrop. The clearest entry to the Chinese open-weight story for a Western listener.
Jordan Schneider's policy + AI fight from inside the China-tech beat. The inside-baseball version for people already tracking GPU export controls and open-model release strategy.
Where to start
3 routes in, one per reader
Nathan Lambert's RLVR Revolution is the patient explainer of why open-weight post-training caught up with closed-lab equivalents.
Reflection AI Secures $2B Funding to Compete with DeepSeek is the sharpest single read on what catching up to the Chinese open-weight pace actually costs.
China's AI Upstarts: How Z.ai Builds, Benchmarks & Ships in Hours is the most concrete view of how a Chinese frontier lab actually operates.
Glossary
Words hosts use without explaining
A model where the trained parameters are publicly released, regardless of whether the training data or code is open. Llama, Mistral, DeepSeek, Qwen are the canonical 2026 references.
The post-training recipe that converged across both open and closed labs in 2025. Train on tasks where the reward signal is checkable (math, code, formal proofs).
The Allen Institute for AI's open foundation model line, designed to be fully reproducible — open weights, open training data, open code. The most academically rigorous of the open-weight stacks.
A research lab releasing open-weight models close in capability to the closed frontier. Mistral, AI2, and the Chinese labs (DeepSeek, Z.ai, MiniMax) are the 2026 examples.
The argument that durable closed-weight advantage now sits in distribution, serving cost, enterprise integration, and product packaging — not raw model quality.