Investors should prioritize Anthropic over OpenAI for long-term exposure, as Claude 3.5 Sonnet is currently winning the critical developer "mindshare" and building a superior product environment. Look for entry points into the "Agentic" shift by backing Horizontal AI tools like Cursor or Perplexity, which focus on user workflow and "harnessing" models rather than just raw processing power. Monitor infrastructure plays that manage AI memory and execution, such as E2B, as these "operating systems" are better positioned to capture value than niche vertical software. Avoid companies that merely "burn tokens" as middlemen; instead, seek out platforms like OpenCrawl that own user context and local data to create high switching costs. While current valuations for top-tier AI labs are high, the real "boiling point" for corporate revenue growth is projected for 2026, making this a critical accumulation phase for patient investors.

This analysis extracts key investment insights from the interview with Yusen, a venture capital investor, regarding the current state of AI, the shift toward "Agentic" workflows, and specific company evaluations.

Anthropic (Claude)

The discussion highlights Anthropic as a major contender that has recently gained significant momentum, particularly in the coding sector.

• Coding Dominance: The guest notes that Claude 3.5 Sonnet (referred to as Cloud Code) has become the preferred tool for many developers, surpassing OpenAI’s current offerings in user experience and efficiency.
• The "Harness" Concept: Anthropic is praised for not just building a model, but a "harness" (the product environment around the model). Their Claude Artifacts and coding capabilities are seen as a "long-time horizon" task success.
• Valuation Perspective:
  • Short-term: May be seen as expensive or fairly valued given the high competition.
  • Long-term (10 years): If they become a primary AI "operating system," current valuations ($20B-$40B range mentioned as context) could look like "cashier" (small) amounts compared to a potential $5T-$10T future.

Takeaways

• Bullish on Product Execution: Anthropic is currently winning the "mindshare" of power users and developers through superior product design (Harness) rather than just raw model size.
• Brand Loyalty: There is a "brand lock-in" effect happening; once developers integrate Claude into their workflow, they are less likely to switch back to OpenAI unless there is a massive performance gap.

OpenAI (ChatGPT)

The sentiment toward OpenAI is currently more cautious, focusing on slowing growth and the challenges of monetization.

• Growth Stagnation: The guest suggests that OpenAI’s DAU (Daily Active Users) and subscription growth have hit a plateau in the last six months.
• Monetization Hurdles: Moving from a simple subscription model to advertising or e-commerce is difficult. The guest notes that even a company as strong as ByteDance (TikTok) struggles to balance innovation with aggressive monetization.
• Model vs. Product: While OpenAI has the "first-mover" advantage, they are currently perceived as being in a "defensive" position regarding their coding products (Codex) compared to Anthropic.

Takeaways

• Risk Factor: High operational costs vs. slowing revenue growth. The "burning token" model requires massive, continuous capital.
• Wait-and-See: The market is waiting for "GPT-5" or a significant "Harness" upgrade to regain the lead in user experience.

The "Agentic" Shift & Harnessing

A major theme of the podcast is that the value in AI is shifting from the Model (the brain) to the Harness (the interface/system).

• Harness as the New OS: The "Harness" (like Cursor, Manus, or OpenCloud) is what allows the AI to perform tasks over a long period without constant human attention.
• Horizontal vs. Vertical: The guest favors Horizontal opportunities. Companies that build tools usable across many industries (like coding or general research agents) are seen as more resilient than those building "Vertical SaaS" for a single niche.
• Agentic Loop: The future is not a chatbot you talk to, but an agent that runs in the background (a "proactive agent") to complete complex tasks.

Takeaways

• Investment Opportunity: Look for companies building the "infrastructure" for agents—tools that manage memory, sandboxes for code execution (like E2B), and cross-platform connectivity.
• Efficiency Paradox: While AI increases productivity by 10x, it hasn't yet translated into 10x revenue for corporations. Investors should look for the "boiling point" where this productivity finally hits the bottom line.

Specific Emerging Mentions

The transcript mentions several smaller or specialized players gaining traction in the developer and power-user communities:

• Cursor: Cited as a prime example of a "Harness" that users love, making the underlying model more valuable through a better interface.
• OpenCloud / OpenCrawl: Mentioned as innovative tools that run locally on a user's machine (Mac), allowing the AI to access local files and context, which creates a stronger "memory" than a standard web-based chatbot.
• Perplexity (AI Search): Noted for its strong brand and user habit formation, making it difficult for competitors to displace.
• Manus / Hermes: Mentioned as part of the new wave of "Agentic" startups that are pushing the boundaries of what AI can do autonomously.

Investment Themes & Risks

• The "2026 Opportunity": The guest hints that 2026 will be a massive year as the current "productivity gains" finally manifest into new business models and products that we cannot yet imagine.
• The "Token" Cost Risk: There is a concern that many AI companies are just "burning tokens" without creating unique value. If a company doesn't have a "Harness" or unique data context, they are just a middleman for the big model providers.
• Human Agency: A unique risk mentioned is the "atrophy of human thought." As we outsource thinking to AI, the value of "out-of-distribution" (original/creative) thinking will skyrocket because AI is currently limited to "within-distribution" (existing) data.

Takeaways

• Focus on "Context": The most valuable companies will be those that own the user's Context (their data, their habits, their specific workflow memory).
• Hardware/Software Convergence: Mention of companies like Aura (health tech) and 0x12 (robotics/mechanical) suggests a growing interest in AI that interacts with the physical world, not just digital screens.