Podcast/Weekly/Oscar & Matt

Human in the Loop

A weekly signal filter for the AI economy. Oscar and Matt sift the week's noise, debate real business ideas, and close with two hot takes you won't hear anywhere else.

Subscribe on YouTube Follow on Spotify Listen on Apple Podcasts podcast.vallyseed.com

Now Playing/EP 21

Frontier AI WITHOUT Nvidia, Ox Alpha was Chinese all along

Sep 1, 2026·85 min·Available on YouTube, Spotify & Apple Podcasts

Live Feed

Weekly·Oscar & Matt

Episodes

Every drop, in order.

Frontier AI WITHOUT Nvidia, Ox Alpha was Chinese all along — episode thumbnail

LatestEP 21

Human in the Loop

Sep 1, 202685 min

Frontier AI WITHOUT Nvidia, Ox Alpha was Chinese all along

Ox Alpha was GLM-5.3-Flash. Z.ai says every request in its anonymous public test ran on Chinese AI chips. That does not end Nvidia's dominance. It does show that a frontier-class model can handle real global traffic without Nvidia hardware. In Episode 21 of Human In the Loop, Oscar Gallo and Matt Wozniak test the claim, separate training from inference, and explain why the model, serving software, network, and chips now have to be judged as one system.

Aug 25, 202678 min

A Mysterious AI Model Just Appeared. Nobody Knows Who Built It.

A free million-token AI model appeared online, takes video, retains your prompts, and has no public developer. So who built Ox Alpha? Oscar and Matt test the claims around the anonymous model now running on OpenRouter — it's free during its preview, accepts text, images, and video, supports tool use, and can return up to 131,072 tokens. Early coding results look strong, but the sample is small, the comparisons are uneven, and no lab has claimed the model.

Aug 18, 202693 min

Opus-Class AI on Your Laptop?

Qwen released a 27-billion-parameter model that fits on a well-equipped laptop and beats Claude Opus 4.6 on selected coding benchmarks. Is local AI ready for real company work? Oscar and Matt separate the benchmark claim from the operating reality — Qwen3.8-27B leads Opus 4.6 Max on SWE-bench Pro and CoWorkBench but trails it on Terminal-Bench 2.1 and GPQA Diamond, and quantized builds run about 18 to 20 GB, which puts 32 GB laptops and desktops in the conversation.

Aug 11, 2026108 min

Meta Launches Muse Code: A 24-Hour Coding Agent

Meta launched Muse Code, a terminal coding agent designed to work across large repositories, coordinate background agents, recover after crashes, and keep going for hours. Is this a better coding system or simply more automation to review? Oscar and Matt look past the model benchmark and examine the operating system around Muse Code — persistent workers in isolated worktrees, a local event log of every model call and approval, and one kernel-optimization run that made more than 1,000 tool calls over as long as 24 hours.

Aug 4, 202663 min

AI Made Engineers Faster. Leadership Fell Behind

AI made engineers faster. It did not make companies faster at deciding what to build. Mike Lyons and Greg Pfister of KaiRise join Oscar and Matt, starting with a $4.3 million voter registration system that shipped on time, on budget, and on scope — and got rejected by its users anyway. When engineering is only 20% of delivery, the bottleneck moves to leadership, planning, and knowing when to stop.

Jul 28, 2026116 min

They Want to Ban Kimi K3

Kimi K3 got hot enough to trigger a U.S. ban debate. If Washington restricts a model you can download and run yourself, is that national security, closed-model protection, or both? Oscar and Matt make the Signal or Noise call on Kimi K3, an OpenAI agent breaching Hugging Face, Claude Opus 5, Meta AI's app actions, and the 50-company open-weights letter, then debate two AI business ideas.

Watch Listen

EP 16

New episodes every week

Don't miss a drop.

Subscribe on your platform of choice and get the weekly signal filter, plus short-form clips between episodes.

YouTube Spotify Apple Podcasts

Human in the Loop is produced by VallySeed