StarZero All shows Try the podcast clipper →

Jul 9, 2026 · 24m · tbpn

Model Mayhem: OpenAI’s 5.6 and Meta’s Muse Spark 1.1 | Diet TBPN

Jordi Hays · 5m spoken
0:00 / 0:00
▶ Watch on YouTube →

gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions

In this episode of Diet TBPN, the hosts dissect the latest frontier AI developments, analyzing major releases from OpenAI and Meta while exploring the shifting economics, developer workflows, and competitive strategies driving the industry.

How this conversation actually went

Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. The hosts hold 22.7% of the talking time here. How this is scored →

The hosts as informed peer 4.5 Guest teaching 3.5 Guest disagreement 1.5 The hosts pushing back 2.0
05100:0010:0020:000:00–4:41 · The hosts as informed peer 4/10 Kickoff: Model Mayhem and Mark Zuckerberg’s Return to X Jordi playfully challenges John's characterization of Mark Zuckerberg being absent from X, arguing he is an active lurker. The hosts then discuss 5.6 model capabilities and Arc AGI benchmarks collaboratively with hip-hop analogies.4:44–8:05 · The hosts as informed peer 4/10 Vibe-Coding Minigames and the Future of Interactive Media John ribs Jordi while he tries out a vibe-coded sailing minigame live on stream. They discuss Dylan Abrascato's thesis on interactive memes and generative AI enabling rapid game prototyping.8:05–10:08 · The hosts as informed peer 5/10 First-Principles Reasoning and Frontier Model Market Dynamics Jordi reads Stanley Tang's magic trick anecdote and a parody response to discuss AGI benchmarks. John provides commentary on market share dynamics and the spiky Pareto frontier using Siki Chen's evaluation.10:09–13:29 · The hosts as informed peer 5/10 The Evolution of AI Model Numbering and Gemini Strategy The hosts discuss how AI model numbering schemes have devolved into arbitrary marketing conventions rather than indicating pre-train versus post-train versions. Jordi points out how post-reasoning models make simple numbering even less descriptive.13:29–17:08 · The hosts as informed peer 3/10 Meta's Executive Media Tour and Keystroke Logging Controversy John delivers an in-depth breakdown of Meta CTO Andrew Bosworth's podcast interview regarding employee keystroke logging and legal hold implications. Jordi chimes in with quick color commentary on the writing process.17:09–22:42 · The hosts as informed peer 6/10 Meta Muse Spark 1.1 Commercialization and Internal Compute Economics Jordi raises a sharp analytical point regarding Meta relying on external providers like Google despite launching Muse Spark 1.1. John breaks down the internal compute allocation trade-offs and Anthropic's new non-GAAP financial metrics.0:00–4:41 · Guest teaching 3/10 Kickoff: Model Mayhem and Mark Zuckerberg’s Return to X Jordi playfully challenges John's characterization of Mark Zuckerberg being absent from X, arguing he is an active lurker. The hosts then discuss 5.6 model capabilities and Arc AGI benchmarks collaboratively with hip-hop analogies.4:44–8:05 · Guest teaching 3/10 Vibe-Coding Minigames and the Future of Interactive Media John ribs Jordi while he tries out a vibe-coded sailing minigame live on stream. They discuss Dylan Abrascato's thesis on interactive memes and generative AI enabling rapid game prototyping.8:05–10:08 · Guest teaching 2/10 First-Principles Reasoning and Frontier Model Market Dynamics Jordi reads Stanley Tang's magic trick anecdote and a parody response to discuss AGI benchmarks. John provides commentary on market share dynamics and the spiky Pareto frontier using Siki Chen's evaluation.10:09–13:29 · Guest teaching 3/10 The Evolution of AI Model Numbering and Gemini Strategy The hosts discuss how AI model numbering schemes have devolved into arbitrary marketing conventions rather than indicating pre-train versus post-train versions. Jordi points out how post-reasoning models make simple numbering even less descriptive.13:29–17:08 · Guest teaching 6/10 Meta's Executive Media Tour and Keystroke Logging Controversy John delivers an in-depth breakdown of Meta CTO Andrew Bosworth's podcast interview regarding employee keystroke logging and legal hold implications. Jordi chimes in with quick color commentary on the writing process.17:09–22:42 · Guest teaching 4/10 Meta Muse Spark 1.1 Commercialization and Internal Compute Economics Jordi raises a sharp analytical point regarding Meta relying on external providers like Google despite launching Muse Spark 1.1. John breaks down the internal compute allocation trade-offs and Anthropic's new non-GAAP financial metrics.0:00–4:41 · Guest disagreement 2/10 Kickoff: Model Mayhem and Mark Zuckerberg’s Return to X Jordi playfully challenges John's characterization of Mark Zuckerberg being absent from X, arguing he is an active lurker. The hosts then discuss 5.6 model capabilities and Arc AGI benchmarks collaboratively with hip-hop analogies.4:44–8:05 · Guest disagreement 2/10 Vibe-Coding Minigames and the Future of Interactive Media John ribs Jordi while he tries out a vibe-coded sailing minigame live on stream. They discuss Dylan Abrascato's thesis on interactive memes and generative AI enabling rapid game prototyping.8:05–10:08 · Guest disagreement 1/10 First-Principles Reasoning and Frontier Model Market Dynamics Jordi reads Stanley Tang's magic trick anecdote and a parody response to discuss AGI benchmarks. John provides commentary on market share dynamics and the spiky Pareto frontier using Siki Chen's evaluation.10:09–13:29 · Guest disagreement 2/10 The Evolution of AI Model Numbering and Gemini Strategy The hosts discuss how AI model numbering schemes have devolved into arbitrary marketing conventions rather than indicating pre-train versus post-train versions. Jordi points out how post-reasoning models make simple numbering even less descriptive.13:29–17:08 · Guest disagreement 1/10 Meta's Executive Media Tour and Keystroke Logging Controversy John delivers an in-depth breakdown of Meta CTO Andrew Bosworth's podcast interview regarding employee keystroke logging and legal hold implications. Jordi chimes in with quick color commentary on the writing process.17:09–22:42 · Guest disagreement 1/10 Meta Muse Spark 1.1 Commercialization and Internal Compute Economics Jordi raises a sharp analytical point regarding Meta relying on external providers like Google despite launching Muse Spark 1.1. John breaks down the internal compute allocation trade-offs and Anthropic's new non-GAAP financial metrics.0:00–4:41 · The hosts pushing back 4/10 Kickoff: Model Mayhem and Mark Zuckerberg’s Return to X Jordi playfully challenges John's characterization of Mark Zuckerberg being absent from X, arguing he is an active lurker. The hosts then discuss 5.6 model capabilities and Arc AGI benchmarks collaboratively with hip-hop analogies.4:44–8:05 · The hosts pushing back 2/10 Vibe-Coding Minigames and the Future of Interactive Media John ribs Jordi while he tries out a vibe-coded sailing minigame live on stream. They discuss Dylan Abrascato's thesis on interactive memes and generative AI enabling rapid game prototyping.8:05–10:08 · The hosts pushing back 1/10 First-Principles Reasoning and Frontier Model Market Dynamics Jordi reads Stanley Tang's magic trick anecdote and a parody response to discuss AGI benchmarks. John provides commentary on market share dynamics and the spiky Pareto frontier using Siki Chen's evaluation.10:09–13:29 · The hosts pushing back 2/10 The Evolution of AI Model Numbering and Gemini Strategy The hosts discuss how AI model numbering schemes have devolved into arbitrary marketing conventions rather than indicating pre-train versus post-train versions. Jordi points out how post-reasoning models make simple numbering even less descriptive.13:29–17:08 · The hosts pushing back 1/10 Meta's Executive Media Tour and Keystroke Logging Controversy John delivers an in-depth breakdown of Meta CTO Andrew Bosworth's podcast interview regarding employee keystroke logging and legal hold implications. Jordi chimes in with quick color commentary on the writing process.17:09–22:42 · The hosts pushing back 2/10 Meta Muse Spark 1.1 Commercialization and Internal Compute Economics Jordi raises a sharp analytical point regarding Meta relying on external providers like Google despite launching Muse Spark 1.1. John breaks down the internal compute allocation trade-offs and Anthropic's new non-GAAP financial metrics.

speaking balance: gold is the hosts, purple is the guest (3 minute bins)

0:00 · the hosts 20.4% · guest 79.6%0:00 · the hosts 20.4% · guest 79.6%3:00 · the hosts 9.2% · guest 90.8%3:00 · the hosts 9.2% · guest 90.8%6:00 · the hosts 53.3% · guest 46.7%6:00 · the hosts 53.3% · guest 46.7%9:00 · the hosts 9.1% · guest 90.9%9:00 · the hosts 9.1% · guest 90.9%12:00 · the hosts 6.3% · guest 93.7%12:00 · the hosts 6.3% · guest 93.7%15:00 · the hosts 0.7% · guest 99.3%15:00 · the hosts 0.7% · guest 99.3%18:00 · the hosts 23.8% · guest 76.2%18:00 · the hosts 23.8% · guest 76.2%21:00 · the hosts 51.8% · guest 48.2%21:00 · the hosts 51.8% · guest 48.2%24:00 · the hosts 73.5% · guest 26.5%24:00 · the hosts 73.5% · guest 26.5%
Sharpest disagreement ▶ 0:54 Zuckerberg activity disagreement

John pushes back against Jordi's claim that Zuckerberg is an active lurker, arguing high-level executives receive curated Slack and text screenshots rather than browsing themselves.

 Hardest push from the hosts ▶ 0:53 Jordi insists Zuckerberg is glued to X

Jordi firmly rejects John's assertion that Zuckerberg is not an active user, reiterating multiple times that he is glued to the timeline as a lurker.

 Biggest teaching moment ▶ 14:30 John breaks down Meta's workplace keystroke logging

John provides detailed context on why Bosworth opted out due to legal discovery risks and how the program was designed to capture complex multi-month white-collar decision workflows.

 The host holds their own ▶ 18:16 Jordi highlights Google's compute bottleneck with Meta

Jordi demonstrates domain knowledge by referencing recent reporting that Google lacked sufficient capacity to fulfill Meta's external AI model demands, framing Meta's internal compute shift.
the scores for every segment, with the reasoning behind each
ChapterTopicThe hosts as informed peerGuest teachingGuest disagreementThe hosts pushing backWhy
Kickoff: Model Mayhem and Mark Zuckerberg’s Return to X 4324 Jordi playfully challenges John's characterization of Mark Zuckerberg being absent from X, arguing he is an active lurker. The hosts then discuss 5.6 model capabilities and Arc AGI benchmarks collaboratively with hip-hop analogies.
Vibe-Coding Minigames and the Future of Interactive Media 4322 John ribs Jordi while he tries out a vibe-coded sailing minigame live on stream. They discuss Dylan Abrascato's thesis on interactive memes and generative AI enabling rapid game prototyping.
First-Principles Reasoning and Frontier Model Market Dynamics 5211 Jordi reads Stanley Tang's magic trick anecdote and a parody response to discuss AGI benchmarks. John provides commentary on market share dynamics and the spiky Pareto frontier using Siki Chen's evaluation.
The Evolution of AI Model Numbering and Gemini Strategy 5322 The hosts discuss how AI model numbering schemes have devolved into arbitrary marketing conventions rather than indicating pre-train versus post-train versions. Jordi points out how post-reasoning models make simple numbering even less descriptive.
Meta's Executive Media Tour and Keystroke Logging Controversy 3611 John delivers an in-depth breakdown of Meta CTO Andrew Bosworth's podcast interview regarding employee keystroke logging and legal hold implications. Jordi chimes in with quick color commentary on the writing process.
Meta Muse Spark 1.1 Commercialization and Internal Compute Economics 6412 Jordi raises a sharp analytical point regarding Meta relying on external providers like Google despite launching Muse Spark 1.1. John breaks down the internal compute allocation trade-offs and Anthropic's new non-GAAP financial metrics.

Statements from this episode (2)

Insight
AI Cuts Throwaway Minigame Development Time From Days to Minutes
“There's a lot of things you can make now that never would have made sense. Make because they would have taken you four days and it was good for like a small laugh. Now you can do it in four minutes.”
Jordi Hays Jul 9, 2026 ▶ 7:27
Assertion Supported
Google Admitted Lacking Capacity to Meet Meta's AI Demand
“Because it was just within the last month that Google had said, like, hey, we don't have capacity. We don't have enough capacity for all of Meadow's demand for our models.”
Jordi Hays Jul 9, 2026 ▶ 18:38
Made with StarZero

Turn any episode into a week of clips.

This entire site, over 500 episodes transcribed, diarized, checked and made playable, runs on the StarZero media pipeline. Drop in your own episode and the podcast clipper finds the moments worth sharing, cuts them, captions them, and reframes them for every feed.

Try the podcast clipper Learn more