Investors should prioritize companies that are "putting Claude at the center" of their operations, specifically targeting enterprise leaders like Salesforce (CRM), Deloitte, and Accenture (ACN) who are early adopters of agentic workflows. Avoid legacy SaaS firms that rely solely on high switching costs for their competitive moat, as AI agents now allow for rapid, low-cost migration between software vendors. Look for significant margin expansion in large-scale organizations like major banks and pharmaceutical companies that are using Claude Code to automate the migration of expensive technical debt from COBOL to modern languages. High-conviction opportunities lie in "AI-native" startups disrupting Legal Services and Material Science, as well as firms utilizing "Frontier" proprietary models rather than lagging open-source alternatives. Monitor the emergence of "multiplayer" AI environments in platforms like Slack (CRM), where proactive agents are expected to drive the next major surge in workplace productivity.

Anthropic (Claude / Claude Code)

The discussion centers on Anthropic, an AI safety and research company, and its breakout software product, Claude Code. The guest, Boris Cherny (Head of Claude Code), highlights how the tool has shifted from a "side project" to a central pillar of the company's business and internal operations.

• Claude Code & Agentic Workflows: Unlike standard chatbots, Claude Code acts as an agent that can interact with a computer's file system, run commands, and iterate on tasks until a goal is reached.
• Model Progression: The rapid adoption in 2026 is attributed almost entirely to improvements in the underlying models: Opus 4.5, Opus 4.6, and the latest Fable (a version of the high-capability Mythos model).
• Internal "Dogfooding": Anthropic reports that 90% to 100% of their internal code is now written using Claude Code.
• Security & Safety: Anthropic utilizes "mechanistic interpretability" to detect "prompt injections" (malicious instructions hidden in data) at the neural level. They claim their model was the only one to resist a week-long hacking competition by external researchers.
• The "Sandbox": To mitigate risks, Claude Code runs in an open-source "sandbox" that restricts the AI's access to specific files and websites, preventing it from "going off the rails."

Takeaways

• Shift in Engineering Roles: The role of the software engineer is moving away from manual typing toward "orchestration." New roles are emerging: Prototypers, Builders, Scalers, and Perfectors (who polish AI-generated output).
• Enterprise Adoption: Major firms like Salesforce, Deloitte, Accenture, Airbnb, and Ramp are integrating these tools. Investors should watch for productivity spikes in companies that successfully "put Claude at the center" of their business processes rather than treating it as a peripheral tool.
• Competitive Moat: Anthropic argues that while "switching costs" (a traditional software moat) may decrease because AI can easily port code between vendors, other moats like network effects and scale economies remain vital.

Software as a Service (SaaS) Sector

The transcript discusses the "SaaSpocalypse"—a market scare where software company valuations were hit due to the fear that AI agents like Claude Code would make traditional software redundant.

• Legacy Code Migration: Claude Code is being used by large banks and organizations to migrate ancient COBOL codebases to modern languages like Java or Rust. What used to take a team of engineers a year can now be done in days for a fraction of the cost (e.g., $150,000 in API credits vs. millions in salaries).
• Disruption of "Switching Costs": AI makes it significantly easier for a business to move its data and logic from one software vendor to another, potentially hurting companies that rely on "vendor lock-in."

Takeaways

• Investment Risk: Companies whose only competitive advantage is high switching costs are at risk.
• Investment Opportunity: Look for "AI-native" startups (like those in Y Combinator) that use agents to disrupt traditional industries like Material Science or Legal Services.
• Efficiency Gains: Large enterprises (Banks, Pharma, NASA) using AI for "code migrations" may see significant long-term savings on technical debt and maintenance.

Artificial Intelligence Themes & Trends

The podcast highlights several broader investment themes within the AI space for 2026.

• The "Harness" vs. The "Model": There is a distinction between the raw AI model and the "harness" (the user interface/terminal). While models are becoming commodities, the safety and integration features of the harness (like Claude Code or Slack integrations) provide the actual business value.
• Computer Use & Visual Interfaces: AI is moving toward "Computer Use," where the agent can view a screen, mock up designs, and interact with apps like a human. This may eventually make traditional hierarchical file systems and complex GUIs (Graphical User Interfaces) less necessary.
• Scaling Laws: The "Scaling Laws" (the relationship between compute, data, and intelligence) continue to hold, and intelligence is currently "accelerating" beyond initial predictions.

Takeaways

• Product Overhang: There is a "product overhang" where models are currently more capable than the software interfaces allow. As interfaces (like Claude Tag or Loops) catch up, expect a surge in AI utility.
• The "Frontier" Advantage: The guest argues against using "open source" models for complex tasks, suggesting that businesses that stay on the "Frontier" (using the most advanced proprietary models) will have a significant competitive advantage as progress continues to move rapidly.
• Multiplayer AI: The future of work involves "multiplayer" environments (like Slack) where AI agents proactively join conversations, query data, and provide solutions without being explicitly prompted.