This Day in AI Podcast

Episode: Why Sam Altman is Scared & Why People Are Giving Up on MCP | EP99.23

Hosts: Michael Sharkey, Chris Sharkey
Date: October 31, 2025

Episode Overview

In this episode, Michael and Chris Sharkey dissect what initially seemed like a quiet week in AI but revealed itself as pivotal due to strategic announcements from OpenAI and notable product developments in the AI ecosystem. The brothers dig into Sam Altman’s recent shift in OpenAI’s strategy, the implications for applications built on top of large AI models, and the existential threats facing foundational AI providers. They also explore the rising tide of specialized AI applications, especially the future of MCPs (Multi-Component Plugins), and cap it off with a humorous but insightful look at the latest in humanoid robotics hype. The episode concludes with a memorable, reflective "sad song" for Greg Brockman.

Key Discussion Points & Insights

1. OpenAI’s Strategic Pivot and Sam Altman’s Message

[00:02–05:49]

- Overview of the OpenAI "State of the Union":

• Sam Altman and another executive (dubbed "the AI version of Elon Musk") take the stage to discuss strategic changes.
• Microsoft is revealed as OpenAI’s single largest shareholder.
• OpenAI signals a transition from focusing on building first-party applications (like ChatGPT and Atlas browser) to enabling a third-party application ecosystem.

“We want to empower people with AI as much as possible and then trust that the process ... of people building better and better things with newer and better tools will continue to go on.”
— Sam Altman, [01:49]

- OpenAI’s Three Pillars:

• Research (to build AGI)
• Product/platform (for ease of use)
• Infrastructure at scale and low cost.

- Shift to Platform Strategy:

• OpenAI plans to act as a foundational AI layer (“AI AGI Cloud Provider”), enabling others to build specialized solutions.

“They’re saying, hey, we don’t want to be out there competing at the application level anymore. We just want all the applications on top of us so we can have the platform, the ecosystem ... be like the Apple tax model.”
— Michael, [03:41]

2. Rise of Specialized AI Applications & Threats to Foundation Model Providers

[05:51–12:49]

- Cursor’s Playbook:

• Cursor releases its own proprietary coding model, Composer, claiming significant performance boosts and lower costs.
• The pathway: build on existing APIs, gather use-case data, then train bespoke models, reduce reliance on incumbents.

“If you build an application that gets really popular ... it’s now becoming possible ... to train a model that can do the things that that application is well known for.”
— Michael, [04:42]

- The Moat Myth:

• Discussion on whether OpenAI and others have any durable moat.
• Open-source models and rapid commoditization mean organizations will not lock themselves into single models or providers.

“I don’t think there’s a real moat around the model anymore.”
— Michael, [09:36]

- User Behavior:

• Organizations use multiple providers and models, optimizing for specific tasks.

“Their teams ... have multiple subscriptions and try different models. ... I don’t really see the point [in locking in].”
— Michael, [10:56]

3. Enterprise, Security, and the MCP Layer

[13:13–16:57]

- Enterprise Adoption Barriers:

• Security, regional data storage, and private deployments are priorities.
• OpenAI lacks the cloud infrastructure and global rollout capacity of giants like Google, Microsoft, and AWS.

- The Power of MCPs:

• MCPs (Multi-Component Plugins) considered by Michael the biggest unlock for enterprise AI — secure, flexible data integrations across disparate systems for smarter decision-making.
• Key question: Will large software providers unlock their platforms via MCPs or resist ceding value to foundational AI clouds?

“Who can provide a platform that allows enterprises to do what you just described securely? I think that is the key.”
— Chris, [15:52]

- Critique of OpenAI’s Product Bloat:

• Release of underwhelming products (AI browsers, video service) distracts from progress on real enterprise needs.
• Anthropic lauded for focusing on modular skills and workflows directly affecting productivity.

4. Platform “Junk Stores,” and Lessons from History

[22:04–28:05]

- OpenAI’s Storefront Vision:

• OpenAI pitches building and monetizing AI apps on their proprietary "store.”
• Hosts draw parallels to past platform ecosystems (e.g., Facebook games boom and bust) — suggesting risk for developers in building on someone else’s platform.

“Zynga ... built a whole business ... on the Facebook sort of SDK. ... And what happened? It doesn’t even exist anymore.”
— Michael, [25:40]

- What Users and Developers Really Want:

• Desire for reproducible, trainable agentic workflows—AI skills that mirror the expertise of trained employees, not flashy interfaces or app stores.

5. Developer Tools and Economics: The Cursor Model

[32:47–37:55]

- Cursor’s Success Story:

• Cursor, by first capitalizing on foundational models (Anthropic and OpenAI), amassed a user base and workflow data, then shifted to their own faster, cheaper model.
• Major productivity boost not through outright replacement but by revolutionizing developer workflows.
• Exemplifies the “build on APIs, learn, then build/buy your own model” playbook now available to others.

“You’ve basically used an available tool at a high cost ... and just accepting that burn ... then they’re like, ‘hey, let’s cut you guys out completely, run our own stuff ... and we just bank money.’ It’s a brilliant business strategy.”
— Chris, [37:00]

6. Market Dynamics and Platform Threats

[38:45–43:56]

- No Loyalty, Only Value:

• Major institutions will keep flexibility: never lock into single vendors; risk of “paying the maximum and not necessarily have the best model in the long run.”

“Why the hell would you ever sign a deal directly with one of the major providers... because something better and cheaper might come along?”
— Chris, [38:45]

- Brand Decay:

• ChatGPT’s lead is tenuous; could be eroded if big tech (Google/Apple) executes well.
• Specialist workflows/applications have defensibility. Altman’s recent acknowledgment of this is a major directional signal.

7. MCPs: The State of Disrepair vs. Real-World Potential

[44:24–65:55]

- User Frustration with MCPs:

• Poor implementations from both OpenAI and Anthropic—unintuitive UI, error-prone setup—fuel negative perceptions.
• Critique: Blaming the protocol, not the implementation.

“These companies have failed to make [MCPs] work well ... everyone blames MCP. Like it’s MCP’s fault as a protocol.”
— Michael, [46:15]

- Actual Utility When Implemented Right:

• Michael shares a story: using Sim Theory’s implementation of MCPs, his AI lawyer assistant (with tool access to email, documents, legal databases) challenged a utility bill—found errors, referenced legislation and previous cases, and ultimately secured a $732 credit.

  “Now you have all the resources they do, and more.”
  — Michael, [62:09]

• Sim Theory credited for making this possible via frictionless, robust MCP integration.

• Many companies’ poor user experience is due to superficial “API factory” approaches, rather than thoughtful tool and workflow design.

“It’s more about the way you work with it that’s important than the thing itself.”
— Chris, [48:32]

- How to Fix MCPs:

• Design tools as intelligent, task-level procedures, not just brute-force API endpoints.
• Inspiration drawn from Anthropic’s “skills,” which encapsulate tool logic and context for smarter, more reliable automation.
• Importance of skills as curated, agentic bundles that enable flexible, reliable, and scalable workflows.

8. AI-Driven Robotics: The “Neo” Pump and Dump?

[68:19–84:43]

- The 1X “Neo” Home Robot Hype:

• Company offers preorders/subscriptions for a humanoid home robot, claiming it can fold laundry, pack/unpack dishwasher, etc.
• Reality: Demos are mostly teleoperated (humans in VR headsets), leading to comparisons with the fake “Mechanical Turk” automaton.
• Major privacy and security fears: allowing strangers to remote view/control in one’s home.
• Hosts compare this to Apple Vision Pro: a “glimpse of the future” but likely headed for the closet once the novelty wears off.

- Human Impact and Bubble Concerns:

• Speculation that, when true general-purpose home robots do arrive, the economic effects will be enormous—just as the “full self driving” demo in Australia led Michael to an “existential crisis.”
• For now, most current humanoid robotics efforts are overhyped and mostly fundraising plays.

“This is just a pump and dump scam. That’s all it is ... None of this matters because [Neo] is just a pump and dump scam.”
— Michael, [77:49]

9. Closing Thoughts: OpenAI, MCPs, and the True Battle

[84:43–86:05]

- Where OpenAI Lost the Plot

• They need to go dark (“Willy Wonka style”), refocus on killer use cases, and deliver technology that truly solves developer and business pain points.
• Winning the race means best-in-class, low-cost, blazingly fast models—otherwise, commoditization will erode any lead.

“Go away, focus on real use cases, and instead of just talking about it, deliver this future.”
— Michael, [85:44]

- Monetization Dilemmas

• The market has been trained to expect cheap, all-you-can-eat AI, but foundational costs are enormous.
• Long-term viability will mean both innovation and sustainable business models.

10. Memorable Moments & Notable Quotes

• On OpenAI’s Real Asset:

  “I think brand recognition and just the fact that so many people have accounts and use it and that if AI is mentioned in, in media, it is chatgpt that is mentioned. ... It’s always ChatGPT. I think that’s what they have.”
  — Chris, [12:49]

• On the Future of AI Workflows:

  “It’s actually the architecture of how you arrange these skills and tools that will lead to, you know, better intelligence later.”
  — Chris, [41:08]

• On Current vs. Promised Capabilities:

  “Actions speak louder than words. Like you’re releasing a browser, you’re releasing some doom scrolling video service. Where’s everything else?”
  — Michael, [16:58]

• On Building on Other Companies’ Platforms:

  “All of those companies ... rose up and some did quite well, but then everything completely disappeared and it sort of goes against the whole open web as well.”
  — Michael, [25:40]

• On the Risk of AI Bubble:

  “What concerns me is Sam Altman single-handedly has set up probably ... the next financial bubble.”
  — Michael, [29:59]

• MCPs in Practice:

  “I fire up the AI lawyer … went through my email, found ... historical power bills ... put together a case ... I sent it [to the company], and then they ... waived the debt ... My point is the power of working MCPs.”
  — Michael, [59:46–62:09]

11. The Greg Brockman "Sad Song"

[88:07–92:59]

The episode closes with a haunting, poetic monologue attributed to Greg Brockman, reflecting on displacement, the passing of AI leadership torches, and calling for conscience over blind acceleration:

“If all our engines roar and rise, who keeps the compass for our skies? ... Remember, guard the human soul. Turn down the glare. Let conscience in, not just to race but learn to win.”
— Greg Brockman (poem narrator), [88:07–92:59]

Timestamps of Key Segments

• OpenAI “State of the Union” / Sam Altman’s pitch: [01:49]
• Cursor 2.0 and the Composer Model: [04:42] & [33:39]
• Economic threats to big AI platforms: [38:45], [41:08]
• Why MCPs are failing in the mainstream: [44:24]
• Michael’s AI lawyer story: [58:03–62:09]
• Neo home robot segment: [68:19–84:43]
• Greg Brockman sad song (poetic monologue): [88:07–92:59]

Tone & Style

The discussion is spirited, irreverent, and peppered with both technical skepticism and tongue-in-cheek hot takes. The hosts maintain a “proudly mediocre” but deeply practical perspective, eschewing academic theory for straight talk about what works, what’s broken, and what the AI hype cycle gets wrong.

Summary for New Listeners

If you haven’t heard this episode, expect:

• Sharp, down-to-earth analysis of major AI provider strategies
• Real-world tales from the application trenches
• Deep skepticism of marketing hype (“AI AGI cloud,” robot butlers)
• Funny, relatable takes on the pain and occasional triumphs of using AI tools today
• A poetic, even moving, meditation on the human consequences of tech progress

Standout quote to close:

“If every workflow wins the war, who tends the soul at the core? ... Let mercy be the reason why.”
— Greg Brockman (poem narrator), [91:57–92:48]

End of summary.