This episode reveals why leading AI developers are strategically distancing themselves from crypto, exploring the real-world battle between Machine Communication Protocol (MCP) and browser-based agents, and what the global energy race means for the future of AGI.
The Strategic Pivot Away from Crypto
- The discussion opens with a candid admission from the speakers about a deliberate strategic shift to separate their AI ventures from any association with crypto. This move is driven by direct feedback from the market, particularly from potential enterprise customers and large developer tool companies who express hesitation or disinterest upon discovering crypto-related branding.
- One speaker recounts how promising conversations with potential customers would halt abruptly once the "crypto" connection was made. This led to a "cleansing" of their website and marketing materials to remove on-chain references.
- Don confirms this sentiment, noting that while their core technology might involve crypto, they have gone to "somewhat of an extreme to remove any of the reference on the website" to avoid alienating mainstream partners.
- The consensus is that the non-crypto space is where commercially viable agentic activity is currently happening. The speakers agree to pivot the podcast's focus accordingly, treating AI as the primary subject and mentioning crypto only when directly relevant.
"If they go to our website and they get like crypto vibes... you know, risks them not engaging."
Strategic Implication: For investors and researchers, this is a critical market signal. The friction between crypto branding and enterprise AI adoption is significant. Projects that can successfully bridge this gap or operate without crypto-centric messaging may have a clearer path to commercialization and partnership in the current climate.
Reassessing Crypto's Role in the AI Stack
- The conversation transitions to a broader reflection on crypto's utility, with the speakers concluding its role is far more limited than previously hyped. They argue that outside of payments and settlement rails for agent-to-agent transactions, most applications of crypto in the AI space remain highly academic and have not demonstrated significant real-world traction.
- One speaker, who previously tried to apply crypto to ownership and data, now believes its primary, and perhaps only, viable use case is payments.
- The AI crypto space is described as being focused on two distinct areas:
- Decentralized Infrastructure: Harvesting data and performing training/inference. It is not yet clear if decentralized architectures will outperform centralized alternatives.
- On-Chain Agent Interaction: Building agents that transact on-chain. This remains largely theoretical.
- Even established crypto sectors like DeFi (Decentralized Finance) are seen as having stalled beyond basic swap and lending primitives, with little meaningful innovation.
Strategic Implication: Investors should critically evaluate the value proposition of "decentralized AI" projects. The most promising near-term applications appear to be those leveraging crypto for specific functions like micropayments between agents, rather than attempting to decentralize the entire AI stack.
The MCP Adoption Problem: A Gap Between Theory and Practice
- A recent trip to San Francisco for a series of AI events provided a stark reality check on the state of agent tooling. The speaker who attended found that while many developer tool companies have launched MCP (Machine Communication Protocol) servers, these servers are seeing almost no real-world usage. MCP is a protocol designed to standardize communication between AI agents and digital tools or services.
- The core issue identified is that most MCP servers are merely "thin wrappers" around existing APIs, offering no new functionality. Developers, therefore, have no compelling reason to switch from using the familiar APIs.
- Dev tool companies themselves seem unsure of the use cases for their own MCP servers, expressing a hope that users will eventually migrate without a clear catalyst.
- The key question becomes: what are the unique use cases that leverage MCP's advanced features, such as bi-directional communication, that are impossible with a standard API?
Strategic Implication: The lack of MCP adoption highlights a critical bottleneck in the agent ecosystem. The value of agent-native protocols will only be realized when they enable functionality beyond what APIs can offer. Researchers and builders should focus on identifying and developing these unique, high-value use cases.
Browser Agents vs. MCP: Is the Browser a "Head Fake"?
- The discussion turns to a LinkedIn article by Boris Wang, which argues that browser-based agents are a "head fake" and that the future lies with agent-native interfaces like MCP. The speakers largely agree, drawing a parallel to the evolution from mobile web browsing to native mobile apps.
- The Mobile App Analogy: Just as early mobile experiences tried to cram websites onto small screens before native apps took over, browser-based agents are seen as a transitional step. The future is likely a more agent-native integration.
- Operational Overload: Agents consuming information via browsers would place a massive load on websites. Businesses have a disincentive to allow this, as agents are not influenced by advertising or branding, providing no economic benefit to the site owner.
- Don summarizes the argument: "If you're going to use agents, you're going to move towards a more agent-native integration flow which would be something like MCP."
Strategic Implication: This suggests a long-term architectural shift away from web-scraping agents toward more efficient, structured communication protocols. Investors should be wary of solutions that rely solely on browser automation, as they may face scalability and economic viability challenges. The future competitive advantage will likely belong to platforms that master agent-native interaction.
The Case for Agent-Native and Generative UIs
- The group explores the deeper technical advantages of agent-native protocols, focusing on bi-directional communication and the potential for proactive, intelligent interaction from the tool's side—something impossible with a static website or API.
- Bi-Directional Communication: Unlike an API where a client only requests data, MCP allows the server (the tool) to query the client (the agent). This could enable a "continuous dialogue," where a dev tool agent could ask a coding agent clarifying questions or suggest better approaches in real-time.
- Proactive Agentic Behavior: A website is static, but a tool with an "agentic" backend could proactively reach out to the user, change the UI on the fly, or even push relevant marketing content.
- A new protocol called AGUI (Agent-to-Generative UI) is mentioned. Developed by Copilot Kit, AGUI aims to standardize how agentic backends interact with and dynamically present information on a front-end, further moving away from static interfaces.
Strategic Implication: The future of user experience in AI is not just chat; it's dynamic, generative UIs that adapt in real-time. AGUI represents an emerging standard in this space. Researchers should track the development of such protocols, as they could define the next generation of human-computer interaction.
The A2A vs. MCP Debate and the Search for Use Cases
- The conversation briefly touches on the distinction between different agent communication protocols, noting that consensus is forming around their specific roles. However, adoption for all protocols remains hampered by a lack of clear, compelling use cases.
- A2A (Agent-to-Agent) is a protocol for communication between different, often untrusted, agents. Its adoption is lagging because there are few production-level, heterogeneous multi-agent systems.
- The emerging view is a separation of concerns:
- MCP: Agent-to-Tool
- A2A: Agent-to-Agent
- AGUI: Agent-to-UI
- The core challenge remains the same for all protocols: identifying killer use cases. As one speaker notes, "We don't even know what to use MCP for... and I think it's like definitely true of like agent to agent as well."
Strategic Implication: The development of a robust multi-agent economy depends on solving the use case problem first. The "Upwork for agents" concept is a powerful vision, but it requires identifying valuable combinations of specialized agents (e.g., a coding agent collaborating with a website analytics agent) to drive the need for A2A protocol adoption.
The Future of AI: The Global Energy Race
- The final segment shifts to the macro-level constraints on AI progress, specifically the immense energy required to train and run frontier models like the anticipated GPT-5. The speakers compare the strategic approaches of the US and China, highlighting energy infrastructure as a critical geopolitical battleground.
- China's Energy Strategy: A speaker, recently returned from China, notes their massive investment in energy infrastructure. This includes a new dam in Tibet projected to produce six times the energy of the Three Gorges Dam and a dominant shift to electric vehicles (EVs) in cities like Shanghai, reflecting a policy of energy self-sufficiency.
- The US Approach: The US is seen as relying on natural gas as a short-term solution while exploring longer-term options like nuclear fusion and small modular reactors. However, it faces significant regulatory hurdles, infrastructure costs, and public opposition that China does not.
- The key takeaway is that while the US excels at innovation (e.g., chip design), China's ability to execute massive, state-directed infrastructure projects may give it a long-term advantage in the raw power needed for AGI.
"The US might not have the energy capacity to continue with moving towards ASI or AGI... but I think the Chinese are going to have it."
Strategic Implication: Energy availability is the ultimate physical bottleneck for AGI development. Investors and researchers must factor energy policy and infrastructure development into their long-term forecasts for the AI sector. The race for AGI is increasingly becoming a race for cheap, abundant, and scalable power.
Conclusion
This discussion highlights a critical pivot in the AI space, where practical enterprise adoption is forcing a move away from crypto-centric branding. The key challenge now is to prove the value of agent-native protocols like MCP over legacy APIs, a step necessary to unlock truly intelligent, scalable agentic systems.
