AI For Humans: Weekly AI News, Tools & Trends

Episode: OpenAI's GPT-5.4 Is a Beast. But Good Luck Staying King.
Date: March 6, 2026
Hosts: Kevin Pereira & Gavin Purcell

Episode Overview

This week’s episode focuses on the release of OpenAI's new GPT-5.4 model, its technical and economic impacts, and the shifting power dynamics in the AI industry—particularly against the backdrop of government contracts and corporate maneuvering. The hosts also delve into major updates from Anthropic and other AI companies, notable AI-powered creative tools, and the broader societal consequences of accelerating AI capabilities.

Key Discussion Points & Insights

1. OpenAI's GPT-5.4 Model: Features & Implications

[00:00–14:41]

• Introduction of GPT-5.4:
  The new model offers substantial improvements in computer use, outperforming the average human and enhancing efficiency in coding tasks, spreadsheet manipulation, presentations, and tool usage.

  • "It is very good at the sorts of things humans do... we are inching ever closer to that ability for AIs to do most of what we do." — Kevin [02:07]

• Naming Conventions:
  OpenAI’s messy naming system persists—GPT-5.3 remaining for the Codex model (coding-focused), while 5.4 is for general use.

• Benchmark Highlights:

  • OS World Benchmark:
    • Average human: ~72%
    • GPT-5.4: 75%
      "On the broad using-a-computer thing, it's better than the average human." — Kevin [05:09]
    • Excels at spreadsheet use and generating presentations.
  • Reduced Hallucinations:
    • 33% reduction in confabulated outputs. "Hallucinations with this model, according to OpenAI's benchmarks again, they're down 33%." — Kevin [09:09]
  • Legal Applications:
    • Scored 91% on Harvey’s Big Law benchmark, making it attractive for legal professionals.

• Efficiency & Cost:

  • Uses fewer tokens and is faster, providing a million-token context window.
  • Despite a nominal cost increase (input: $1.75 → $2.50/million tokens), improved token efficiency means real-world usage is cheaper.
  • "Reduced token usage by 47% with no accuracy loss." — Kevin [11:46]
  • "Dan Shipper... is comparing it to Opus 4.6 and he says it's half as much." — Gavin [11:49]

• No Wall in Sight:

  • OpenAI researchers affirm rapid, ongoing progress.
  • "We see no wall and expect AI capabilities to continue to increase dramatically this year." — Noam Brown (quoted by Gavin) [12:17]

2. AI and the Pentagon: High-Stakes Industry Drama

[14:41–26:03]

• Anthropic’s Pentagon Trouble:

  • The Pentagon demanded the removal of two "red lines": no autonomous weapon targeting and no mass US-citizen surveillance.
  • "Our present government did an abrupt about-face and basically said, remove these two red lines. Or else." — Kevin [18:01]
  • Anthropic refused, risking massive contract loss and industry blacklist.

• Dario Amodei’s Stand (Anthropic CEO):

  • "The red lines we have drawn, we drew because we believe that crossing those red lines is contrary to American values." — Dario Amodei (clip) [19:28]

• OpenAI Steps In:

  • OpenAI (Sam Altman) negotiated with the government and appears more compliant, leading to accusations of being mendacious and too accommodating.
  • Industry faces potential government intervention or nationalization pressures—echoes of a new "nuclear race" for AI.
  • "If a private company has the ability to say yes or no based on a technology, then that does become a very big lever for nationalizing that private company, which means that the government takes control." — Gavin [22:56]

• Anthropic’s Rise Amid Turmoil:

  • Despite losing a big military contract, Anthropic's model Claude is surging in popularity and revenue, catching up with OpenAI.
  • "Claude has rocketed to the number one app in the App Store, which it has never been before. Katy Perry is now a Claude max user..." — Gavin [24:36]
  • "Anthropic is making about $19 billion a year right now... OpenAI just said, well we're making 25 billion a year now." — Gavin [25:01]

3. Rapid AI Advancements: Claude Code & Cloud Rotting

[27:06–30:18]

• "Claude Rotting":

  • Gavin's newly coined term for spending excessive time on Claude Code projects — a nod to “bed-rotting” memes.
    • "Claude rotting is basically... spending too much time doing Claude code and not taking care of yourself." — Gavin [27:19]

• Claude Code as Daily Driver:

  • Both hosts now rely on Claude for coding, productivity, and creative projects, including building site features and automating personal workflows.
  • "Claude is my daily driver, full stop. I've got multiple terminal windows open that are running instances of Claude code. My Claude bots are all powered by Claude code." — Kevin [28:39]

• Voice Mode & Skills Growth:

  • New features and increased "pair bonding" with Claude for productive tasks.
  • "Voice mode is coming to cloud code. There are a bunch of new cloud code skills that are coming." — Gavin [28:07]

4. AI Video Tools: Generative Media Explosion

[30:18–42:59]

• Kling Motion Control 3.0:

  • AI video puppeteering; transforms performances onto any character with high fidelity, including nuanced facial expressions and lighting.
    • "You can puppet a video or add a look to a video... it does a very good job of it." — Gavin [30:39]
    • Not perfect with hands/objects but impressive for face/voice transfer.
    • McDonald’s CEO viral clips used as fun benchmarks.

• Grok Imagine Extend:

  • Allows users to extend video clips using AI; works, but still not flawless—generational drift in faces over longer generations.
    • "Grok Extend allows you to take a video and basically extend on the end of it and add stuff on." — Gavin [35:40]

• Sea Dance (Seedance) & Google NotebookLM Video Summaries:

  • Seedance offers more consistent video generation and audio.
  • NotebookLM integrates cinematic explainer videos—potential to disrupt educational or "faceless" YouTube channels, though still early and “AI-feeling.”
    • "When you watch these, it feels like an AI made these things..." — Gavin [40:25]
    • "The number of times I have gone to NotebookLM to digest big topics lately and generated like... a podcast is non-zero." — Kevin [41:34]

5. AI in Hollywood: Ben Affleck’s Secret Startup & Job Automation

[42:59–45:29]

• Ben Affleck’s Inter Positive Acquisition:
  • Affleck’s AI startup, focused on advanced post-production tools, sold to Netflix.
    • "Ben and did this company so that he could... use AI tools to help productions shoot better and more efficiently." — Gavin [44:16]
    • Based on “enhancing” rather than “replacing” human creative work—but possible implications for job loss in production and post.

6. Privacy Tools: Counteracting AI Surveillance

[46:11–48:06]

• Spectre 1 (AI Audio Jammer):

  • Prototype device claims targeted audio jamming to prevent unwanted AI eavesdropping—an anticipated response to ubiquitous, always-listening wearables.
    • "I see a future for devices like this, especially as everybody's got the meta Ray Bans coming around or, you know, whatever Apple's going to release." — Kevin [47:17]

• Wi-Fi Surveillance Risks:

  • Discussion of real-world methods for “seeing through walls” by analyzing WiFi signal perturbations.

7. Lightning Round: Creative AI Projects & Tools

[48:46–57:04]

• The Shape Store (Virality & AI Copyright):
  • Viral TikTok series “Shape Store” — playful, creative example of AI-generated content, challenging copyright boundaries.
    • "This is a really cool thing... this is a creative voice with AI tools put together." — Gavin [50:30]
• Real-Time War Visualization (Billowal Seat):
  • Robust interactive visualization of war data, productized from solo “vibe coding” efforts.
• Infinite Favicons (Joseph Jojo):
  • Mesmerizing “favicons” (website icon) browser.
• Blizzain’s Open-Source AI Music Video Maker:
  • Upload any track; get an autogenerated music video—demonstrating the accessibility of AI-powered creativity.

8. Bonus: AI Cover Songs, Personal Projects, and Final Thoughts

[54:54–End]

• AI Cover Song Experiments:
  • Kevin’s ongoing obsession with AI-generated music covers, including targeting lyrics and vocal stylings with prompt engineering and Claude bots.
    • "Rock and metal bands doing nursery rhyme covers. So you'll get System of a Down doing the Wheels on the Bus. You'll get Rage Against the Machine doing Humpty Dumpty..." — Kevin [57:04]

Memorable Quotes & Moments

• On Naming Chaos:
  "If anything, that just makes me feel so much better about Skynet not happening tomorrow if they can't get the names right." — Kevin [02:39]

• On Breakneck Progress:
  "We might be another algorithmic breakthrough or three away from unlocking this AGI massive superintelligence future." — Kevin [08:34]

• On AI in the Military:
  "Could you imagine the Manhattan Project taking place in public with like three or four, let's say 12 different companies going after it and TikTok exists at the same time?" — Kevin [26:03]

• On Claude’s Appeal:
  "You have the twinkle back in your eye." — Kevin [28:36]
  "Claude is my daily driver, full stop." — Kevin [28:39]

• On AI Video Comedy:
  "It looks like it's like a dog with an Instagram with the tongue hanging out of the mouth." — Kevin (on Kling video tests) [33:21]
  "That's like a Left 4 Dead enemy when it gets agitated." — Kevin [34:32]

• On AI Copyright and Creativity:
  "This is a creative voice with AI tools put together... this is a really cool thing." — Gavin [50:30]

Timestamps for Important Segments

• 00:00–06:30: Introduction to GPT-5.4—improved computer use, cost efficiency, benchmarks
• 09:07: Discussing reduced hallucinations and legal use-cases
• 14:41–18:52: The Pentagon, Anthropic, and red line controversy
• 19:28: Dario Amodei (Anthropic CEO) CBS interview clip
• 22:10–26:03: OpenAI’s government deal, nationalization hypotheticals
• 27:06–30:18: “Claude Rotting”—working on Claude Code projects
• 30:18–39:27: Kling, Grok, Seedance, and other AI video tools
• 42:59–45:29: Ben Affleck's AI company acquired by Netflix
• 46:11–48:06: Audio jamming device “Spectre 1” and WiFi-based surveillance
• 48:46–51:16: Viral AI projects (Shape Store, Infinite Favicons, AI music videos)
• 54:54–57:04: Kevin’s AI music cover experiments and philosophical reflections
• 57:55–End: Show closing, links to podcast extras, calls to support the show

Final Thoughts

This episode emphasizes how quickly AI is evolving, both in technical capability and societal impact. The hosts maintain their signature mix of humor, skepticism, and excitement while breaking down potentially world-altering trends—from coding and creative arts to military contracts and privacy tech. Listeners come away with not just an update on the state of AI, but also a sense of the cultural and ethical negotiations that accompany frontier technology.

For full demos, code snippets, and curated links, visit aiforhumans.show/pod.