This Day in AI Podcast – Episode 99.27

ChatGPT is Dying? OpenAI Code Red, DeepSeek V3.2 Threat & Why Meta Fires Non-AI Workers
Hosts: Michael Sharkey (A), Chris Sharkey (B)
Date: December 4, 2025

Overview

In this episode, Michael and Chris dive deep into the current "Code Red" scenario at OpenAI, dissecting the decline in ChatGPT's market dominance amid mounting competition from models like DeepSeek 3.2, Gemini 3, and others. They candidly discuss the shifting AI landscape in both enterprise and consumer sectors, the challenge of maintaining AI fluency in the workplace, and the broader implications for education and skills in a rapidly changing world. With the duo’s trademark blend of technical curiosity, humor, and healthy skepticism, the Sharkey brothers explore what OpenAI needs to do to reclaim its crown, the cost dynamics facing AI startups, and the cultural shifts AI is driving in business and education.

Key Discussion Points and Insights

1. OpenAI's "Code Red": Competitive Pressures & Waning Hype

• OpenAI's Wobble

  • OpenAI has declared a “code red” internally as reports show a 6% market share loss this year ([00:02]).
  • Efforts to debut ads have been delayed, raising concerns about ChatGPT’s commercial future.
  • "I think their moat was that allure of having the best models and the hype around it. ...But this year, I don't know, it's crickets." — Michael [02:09]

• Loss of Edge

  • Other models like Gemini 3, DeepSeek, and Nano Banana are proving more compelling in specific tasks and are being favored, even by mainstream users.
  • Chris describes seeing a school principal embracing alternative tools (Suno, VO3) rather than sticking to ChatGPT, suggesting real users just want the best tool, not brand loyalty ([00:43]).
  • "People just want to use frontier models ... Whatever is best for the job, that's what they pick." — Chris [18:16]

• Failure to Deepen Consumer Engagement

  • ChatGPT hasn't become indispensable: “Every time you go to ChatGPT, it’s like a Tinder date. You go there for kind of one thing... This is not a relationship tool.” — Michael [07:52]

• Vulnerability of Fashion Status

  • Chris emphasizes that generic, fashion-driven AI use (i.e., not deeply integrated into workflows) is easily displaced by competitors offering similar tools for free or with fewer annoyances, like ads ([04:26], [08:43]).
  • "Nothing is going to piss off the average AI user more than injecting ads into the responses from your AI models." [08:43]

2. Business Model & Enterprise Challenges

• Reliance on Consumer Revenue

  • OpenAI is at a disadvantage compared to cloud providers; they need large contracts to manage hardware costs but mostly serve consumers who can churn easily ([09:54]).
  • Enterprises are wary due to data privacy issues, especially since a mistaken checkbox can let OpenAI train on company data—a dealbreaker for many serious clients ([05:00]).

• Need to Refocus

  • The duo agree OpenAI must get back to delivering the best, most accessible models to regain their lead.
  • "This code red ... should be about building models that people actually are climbing over the walls to use because they're notably better." — Michael [14:19]

• Comparisons to Competitors

  • Google's inclusion of Gemini everywhere inflates engagement metrics, while models like DeepSeek offer open-source, self-hosted options that are far cheaper and more controllable.
  • “...all the startups that are building around AI are using Chinese models. ...DeepSeek v3.2, I think, further erodes for both OpenAI and Anthropic in a lot of ways because it's so much cheaper to use.” — Michael [34:54]

3. DeepSeek 3.2: The Open-Source Threat

• Performance & Cost

  • DeepSeek 3.2 is open source, can be self-hosted, and comes close to top-tier models in benchmarks. The combined cost of DeepSeek for inference and image synthesis (using Nano Banana) is dramatically lower than OpenAI or Anthropic’s fees ([34:46]).
  • "If you did like a blind test on me with this, I probably wouldn't have noticed ... It's pretty amazing what it can do." — Chris [32:26]

• Implications for Startups

  • Startups building on OpenAI’s APIs face poor gross margins, whereas open source models can flip the economics. Only once use cases are validated does it make sense to move to self-hosting for cost control ([38:03]).
  • Privacy is often better with open-source Chinese models (if hosted locally), countering Western data-fear narratives ([36:36]).

4. AI Fluency as a Workplace & Educational Imperative

• Meta's New Direction

  • Meta (Facebook) will begin formally grading employees on AI skills starting in 2026, pushing towards an “AI native culture” and reflecting an industry-wide expectation for AI fluency ([40:28]).
  • “Workers or employees that are not adopting AI are starting to actually fall behind. They're starting to seem like dinosaurs.” — Michael [41:59]

• Blunt Evolution in Job Requirements

  • Non-AI fluent workers are increasingly inefficient: “If it's taking this person more than an hour, they don't know how to use AI basically ... I'd fire them immediately or retrain them obviously.” — Chris [42:40]

• Broader Disruption

  • The value of traditional higher education is called into question as students weigh debt burdens against the essential nature of AI skills.
  • “College isn't preparing students for today's job market. ...This and the debt burden will create a reckoning for higher education.” — Michael, citing Zuckerberg [47:09]
  • Universities need to integrate “AI fluency” into every degree to remain relevant ([48:13]).

• Pragmatic Adoption

  • The brothers discuss positive examples (University of New England, Australia) already moving to embed AI across curricula ([48:44]).
  • Comparison to the internet’s early adoption: "Now it's getting to a point where a lot of this stuff is just overnight out of date." — Michael [47:09]

• Resistance and Backlash

  • Some backlash remains, with skeptics viewing AI-driven tools as “slop,” but workplace and school trends are clearly advancing AI integration ([50:15]).

5. Entertaining Moments & Analogies

• Notable Analogies

  • “Every time you go to ChatGPT, it's like a Tinder date... This is not a relationship tool.” — Michael [07:52]
  • Describing AI tool switching: “You just know which app to use for what purpose and what app yields the better result.” — Michael [21:39]

• Personal & Family Anecdotes

  • Chris shares how even his technically-inclined, retired father-in-law notices when models (e.g., Gemini Pro 2.5) "get dumber," proving average users’ awareness ([20:03]).
  • Schools acknowledge AI’s permeation; Chris jokes about his son’s school banning their own Simtheory app, seeing it as a “badge of honor” ([54:42]).

• Humorous Digression: Songs & AI Art

  • The brothers reference their own AI-created Spotify chart-toppers, wisecracking about getting them played on radio ([56:25]).

• WTF of the Week: Tesla’s Optimus Robot

  • They describe a video of the Tesla humanoid robot and debate its “running” prowess with characteristic banter ([58:05]).

Notable Quotes with Timestamps

• “There’s rumors that there’s four new models going to drop before the end of the year, but … these are panic tunes of the model or benchmark, like benchmark maxing models.” — Michael [02:09]
• “Nothing is going to piss off the average AI user more than injecting ads into the responses from your AI models.” — Chris [08:43]
• “If OpenAI cut us off today, I probably wouldn’t even fight to get it back.” — Chris [16:33]
• “All of the startups that are building around AI are using Chinese models… DeepSeek v3.2, I think, further erodes for both OpenAI and Anthropic in a lot of ways because it’s so much cheaper to use.” — Michael [34:54]
• “Workers or employees that are not adopting AI are starting to actually fall behind. They’re starting to seem like dinosaurs.” — Michael [41:59]
• “Universities need an AI component to basically all of their degrees … that has, like you say, AI fluency attached to it.” — Chris [48:13]
• “It's like when the Internet was a thing, like people like, ‘I don't do the Internet’… the same group of people are like that with AI today.” — Michael [48:56]
• "It’s all a reputation thing. It’s like, can you motivate yourself to stick with Deep Seek for more than a day? ...It’s a vibe thing." — Chris [59:29]

Timestamps for Key Segments

• OpenAI’s "Code Red" and Competitive Threats: [00:02]–[11:06]
• Business Model, Ads, and Enterprise Vulnerabilities: [09:51]–[14:19]
• Blind Loyalty vs Best Tool for the Job: [13:20]–[22:54]
• DeepSeek 3.2 and Open-Source Model Disruption: [31:22]–[39:34]
• AI Fluency in Meta and the Workforce: [39:34]–[48:28]
• Education, Universities, and AI Integration: [48:28]–[54:13]
• School Anecdotes, Bans, and Student Adoption: [54:13]–[56:25]
• AI Songs Banter: [56:25]–[57:55]
• Tesla Optimus "WTF of the Week": [58:05]–[59:53]

Conclusion

Michael and Chris wrap up by acknowledging that while OpenAI remains a major player, its dominance is no longer assured without significant innovation. The episode is a lively, critical exploration of the AI ecosystem’s fast-evolving nature, illuminating not only the technical arms race but the profound societal ramifications on work and learning. The hosts’ accessible, irreverent approach brings clarity and relatability to a complex, rapidly shifting landscape.

For more, check out the full episode and join the discussion about the real, messy, but very real world of living and working with AI.