Why Meta Killed the Metaverse (And is Failing at AI)
Meta's Metaverse is dead — billions burned, Horizon Worlds shut down, and Zuckerberg still can't ship a competitive AI model. Hunter and Daniel break down Meta's 20% layoffs, their failed VR-to-AI pivot, and why the company that renamed itself for the future keeps getting lapped by Google, OpenAI, and Anthropic.
First up: Hunter confesses to running Claude Max in "dangerously skip permissions" mode — and his computer might be infected. Right on cue, LiteLLM (an open-source package half the AI industry depends on) got hit with a supply chain attack that tried to steal every API key and password it could find.
Then it's Meta's autopsy. Horizon Worlds is dead, and Hunter and Daniel revisit their own failed attempt to podcast inside the Metaverse. Why did VR Chat crush Meta's billion-dollar platform? Hunter's theory: you can't build a product for seven-year-olds and forty-seven-year-olds at the same time. With 20% of the company getting laid off, Meta is betting everything on AI — but their models keep underperforming. Llama 4 disappointed. The rumored "Avocado" model supposedly barely matches what competitors shipped a year ago.
Meanwhile, open-weight models from China are eating Meta's lunch. Kimi 2.5 and Qwen 3.5 are running locally on consumer hardware and rivaling the best closed models. Hunter's running Qwen 3.5 on his MacBook Pro and says the chat experience is indistinguishable from Claude or ChatGPT — at least for non-coding tasks. Could the average person ditch their AI subscriptions and go fully local? Almost, but your mom probably isn't installing LM Studio anytime soon.
The episode wraps with the "Pirate and Architect" theory — a vision where vibe-coding pirates ship features at the speed of thought and senior architects clean up behind them. Is this Meta's future? Is it everyone's future? And should Zuckerberg just give up on foundational models and use Gemini like Apple?
⏱️ CHAPTERS
0:00 Intro
2:14 Unprotected Claude Sessions
4:16 LiteLLM Got Hacked
6:12 Meta Killed the Metaverse
8:45 Meta's 20% Layoffs
10:55 Why Meta Can't Build Good AI
13:35 Open-Weight Models Are Here
18:47 Could Your Mom Run Local AI?
24:37 Pirates & Architects
29:26 The Twitter/X Playbook
34:14 That's the Whole Conclusion
⚡ Listen now & get self-aware before your tools do.
🎧 Listen on Spotify: https://open.spotify.com/show/3EcvzkWDRFwnmIXoh7S4Mb?si=3d0f8920382649cc
🍎 Subscribe on Apple Podcasts: https://podcasts.apple.com/us/podcast/they-might-be-self-aware/id1730993297
▶️ Subscribe on YouTube: https://www.youtube.com/channel/UCy9DopLlG7IbOqV-WD25jcw?sub_confirmation=1
📢 Engage
Confess in the comments: are you running your AI tools in YOLO mode right now? No judgment. (Okay, a little judgment.)
New here? Subscribe for twice-weekly AI chaos.
🧠 They Might Be Self-Aware — but are we?
#AI #Meta #Metaverse
First up: Hunter confesses to running Claude Max in "dangerously skip permissions" mode — and his computer might be infected. Right on cue, LiteLLM (an open-source package half the AI industry depends on) got hit with a supply chain attack that tried to steal every API key and password it could find.
Then it's Meta's autopsy. Horizon Worlds is dead, and Hunter and Daniel revisit their own failed attempt to podcast inside the Metaverse. Why did VR Chat crush Meta's billion-dollar platform? Hunter's theory: you can't build a product for seven-year-olds and forty-seven-year-olds at the same time. With 20% of the company getting laid off, Meta is betting everything on AI — but their models keep underperforming. Llama 4 disappointed. The rumored "Avocado" model supposedly barely matches what competitors shipped a year ago.
Meanwhile, open-weight models from China are eating Meta's lunch. Kimi 2.5 and Qwen 3.5 are running locally on consumer hardware and rivaling the best closed models. Hunter's running Qwen 3.5 on his MacBook Pro and says the chat experience is indistinguishable from Claude or ChatGPT — at least for non-coding tasks. Could the average person ditch their AI subscriptions and go fully local? Almost, but your mom probably isn't installing LM Studio anytime soon.
The episode wraps with the "Pirate and Architect" theory — a vision where vibe-coding pirates ship features at the speed of thought and senior architects clean up behind them. Is this Meta's future? Is it everyone's future? And should Zuckerberg just give up on foundational models and use Gemini like Apple?
⏱️ CHAPTERS
0:00 Intro
2:14 Unprotected Claude Sessions
4:16 LiteLLM Got Hacked
6:12 Meta Killed the Metaverse
8:45 Meta's 20% Layoffs
10:55 Why Meta Can't Build Good AI
13:35 Open-Weight Models Are Here
18:47 Could Your Mom Run Local AI?
24:37 Pirates & Architects
29:26 The Twitter/X Playbook
34:14 That's the Whole Conclusion
⚡ Listen now & get self-aware before your tools do.
🎧 Listen on Spotify: https://open.spotify.com/show/3EcvzkWDRFwnmIXoh7S4Mb?si=3d0f8920382649cc
🍎 Subscribe on Apple Podcasts: https://podcasts.apple.com/us/podcast/they-might-be-self-aware/id1730993297
▶️ Subscribe on YouTube: https://www.youtube.com/channel/UCy9DopLlG7IbOqV-WD25jcw?sub_confirmation=1
📢 Engage
Confess in the comments: are you running your AI tools in YOLO mode right now? No judgment. (Okay, a little judgment.)
New here? Subscribe for twice-weekly AI chaos.
🧠 They Might Be Self-Aware — but are we?
#AI #Meta #Metaverse
Loading content...
Read the full transcript
Matterfact has every podcast on Apple Podcasts and Spotify: 700K+ channels, 120M+ episodes.
The candid version is on a podcast,
not the earnings call.
Matterfact turns the conversations your coverage is already having into a searchable, citable record — across every episode, not just this one.