In a recent episode of the “20VC with Harry Stebbings” podcast, Instagram co-founder and Anthropic CPO Mike Krieger discussed where value will be generated in an AI-driven world and the challenges and opportunities facing startups and established companies alike.
Differentiated Strategies for Success
Krieger emphasized that the most valuable opportunities in the AI landscape lie in areas with differentiated go-to-market strategies, specialized industry knowledge, or exclusive data access. He pointed out that sectors like finance, law, and healthcare offer durable opportunities for value creation, requiring significant upfront effort that goes beyond short-term projects.
AI’s Impact on Vertical SaaS Companies and Startups
According to Krieger, both existing vertical SaaS companies and new startups can benefit from the rise of AI. However, they face distinct challenges:
- Established companies must carefully integrate AI to meet customer expectations without overpromising.
- Startups need to balance painting a vision of the future with delivering tangible value quickly, especially when model quality heavily influences product performance.
Krieger advised startups to explore the capabilities of current AI models and aggressively test new ones to capitalize on advancements.
The Model Layer: Talent, Focus, and AI Partnership
Krieger identified three key areas for long-term investment in the foundation model layer:
- Talent: Attracting and retaining top AI talent is crucial for staying at the frontier of innovation.
- Focus and Model Characteristics: Developing unique model characteristics and focusing on specific verticals, such as coding, can lead to significant advantages.
- AI Partnership: Offering AI partnership rather than simply providing AI models is essential for building strong, lasting relationships with companies.
Overcoming Blockers and Measuring “Vibes”
Krieger noted that a major obstacle to AI progress is creating training environments that accurately reflect real-world challenges. He also highlighted the importance of measuring and evaluating the qualitative aspects of AI, such as character and “vibes,” which are currently difficult to assess.
Model Quality vs. Product UX
Krieger stressed that model quality and user experience are inseparable. He argued that designers and product managers need to design products around non-deterministic systems, considering factors like prompting and model evaluation as integral parts of the product design process.
The Speed of Innovation and Product Marketing
With the rapid pace of AI releases, Krieger acknowledged the challenges of product marketing. He emphasized the need for agility and adaptability in responding to new developments and maintaining a long-term perspective on progress.
Brand Matters
Krieger believes that brand identity is crucial, especially on the consumer front. He drew parallels to social networks, where format, audience, and “vibes” contribute to a product’s unique appeal.
The Role of Open Source and Distillation
While open-source models have their place, Krieger expressed concerns about uncontrolled model distillation due to national security implications. He also emphasized the importance of commercializing AI innovation to sustain long-term advancements.
China’s AI Capabilities
Krieger cautioned against underestimating China’s AI capabilities. He noted that China has demonstrated creative thinking and problem-solving at scale, particularly in the development of super apps and other innovative products.
The Future of Work and AI’s Impact on Coding
Krieger believes AI will significantly change the role of software developers. He envisions a shift towards more multi-disciplinary roles, with developers focusing on what to build and delegating implementation to AI models. Code review will also evolve, with developers primarily evaluating AI-generated code.
Read more: Mike Krieger: Why Software Engineers Need to Embrace AI – or Risk Obsolescence
Balancing Consumer and API Businesses
Krieger discussed the trade-offs between building consumer-facing products and API businesses. While first-party products offer valuable learning opportunities and brand-building potential, APIs provide broader distribution and exposure.
Quick Fire Round Highlights
In a rapid-fire Q&A, Krieger shared the following insights:
- OpenAI has been faster at shipping v1s.
- Anthropic excels in personality and cohesive features.
- He respects OpenAI for balancing product development and a widely used API.
- He would rebuild Anthropic’s product to simplify the information architecture.
- He now recognizes the importance of investing in first-party products.
- He is concerned about the lack of discernment and privacy in increasingly capable AI models.
- He sees potential for AI to increase longevity by accelerating drug discovery and clinical trials.
FAQ Inspired by the Podcast
- Where will value be generated in an AI-driven decade? Value will be generated in areas with differentiated go-to-market strategies, specialized industry knowledge, or exclusive data access, particularly in complex sectors like finance, law, and healthcare.
- How should startups approach building products in the rapidly evolving AI landscape? Startups should explore current AI model capabilities, aggressively test new models, and focus on delivering tangible value quickly while painting a vision of the future.
- What are the key areas for long-term investment in the foundation model layer? The key areas are attracting and retaining top AI talent, developing unique model characteristics with a specific focus, and offering AI partnership beyond just providing models.
- What is the biggest blocker to AI progress? The biggest blocker is creating training environments and evaluations that accurately mirror real-world challenges and complexities.
- How will AI change the role of software developers? AI will shift the focus towards more multi-disciplinary roles, with developers concentrating on what to build, delegating implementation to AI models, and primarily evaluating AI-generated code.
