GPT-5, Rotary Phones & the Month of Adaptability

🎭 Opening scene: chaos in L.A., weird calls, and weirder updates

Sentiment: Playful, mildly frazzled, high camaraderie

Kyle kicks things off broadcasting from a hotel business center in L.A., wrestling with flight panic, Wi-Fi threats, and a hairdressing consult happening nearby. The room quickly fills with the week’s obsession: GPT-5 rollout chaos. People are half-bragging, half-lamenting their random access patterns, some have it on mobile only, others lost it overnight, and a few are living in a bizarre limbo of partial features.

Quotable chuckle:

“2025 gave a whole new meaning to Model.” — Gareth Hood


🤖 GPT-5 hits… and breaks stuff

Sentiment: Excited, suspicious, battle-ready

Kyle recaps the launch: unified models, “thinking” mode, advanced voice everywhere (even for custom GPTs), and free-tier access until credits run out. But access feels unstable, accounts rolling back, features disappearing, custom GPTs breaking.

Highlights & gripes:

  • It’s slower than GPT-4, possibly due to server strain.

  • Editing a single part of an AI-generated image without changing the rest? Now possible.

  • Prompt adherence for image generation is better than before.

  • Long-time “model pickers” miss manual control.

Brilliant quote:

“Adaptability is the word for the rest of August — assume everything’s going to break.” — Kyle


🛠 Connectors, agents & deep reasoning… kind of

Sentiment: Curious but confused

The group tries to pin down what “connectors” actually are; basically app integrations (think Gmail, Google Calendar), but the rollout is clunky. Agents, introduced barely a week ago, are already overshadowed by GPT-5 hype. Interface inconsistencies mean some can’t even find the buttons.

Quotable moment:

“I feel like Dora the Explorer all over again.” — Kris Huson


💡 Trust, memory & personality shifts

Sentiment: Intrigued, a little betrayed

Some notice GPT-5 feels more trustworthy and less prone to hallucinations.

Others mourn the personality changes in their favorite GPTs, “like you lobotomized my best friend.”

Memory quirks, project errors, and weird self-identification (“I’m GPT-4.5”) fuel speculation that we’re talking to an orchestration layer rather than a brand-new core model.


🎨 QR codes, beatbox apps & musical AI

Sentiment: Amused, impressed

QR code generation is hit-or-miss, pretty ones often don’t scan, plain ones work fine. HT demos a beatbox app built in one shot via GPT-5 (“formatted for mobile”), while others explore ElevenLabs’ new music tool, Producer AI, and creative copyright angles for AI-generated songs.


📈 Industry tea & competitive context

Sentiment: Cautiously analytical

Debate sparks over whether GPT-5 represents a huge leap. Some point to shrinking performance gaps with Anthropic and Google; others argue the real test is how long competitors take to catch up. Kyle notes OpenAI may have prioritized real-world usage over benchmarks — meaning new prompting styles might unlock hidden performance.


🎯 Takeaways

  1. Skill of the month: Adaptability — nothing is stable, assume relearning.
  2. Connectors could be powerful once they work consistently.
  3. Trust & memory shifts matter as much as speed or accuracy.
  4. Creative play (beatbox apps, live lyric iteration) reveals GPT-5’s potential beyond productivity.
  5. The human factor — attachment to AI personalities — isn’t being measured but may define user loyalty.

🔮 Summary

Vibe: Playful, slightly chaotic, experimentally optimistic

Biggest shift: From model-switching control to a unified “thinking” AI

Standout moment: Live beatbox app demo from a single prompt

Most quotable line: “Adaptability is the word for the rest of August — assume everything’s going to break.”