Instagram Co-founder Mike Krieger on the Shift from Model Competition to Product UX: What It Takes to Win in the AI Era

"The evolution of AI is faster than imagined, with performance improving significantly every few months."
Mike Krieger, who co-founded Instagram and currently serves as CPO at the AI lab Anthropic, has been at the forefront of the tech industry. In this article, we explore his insights on "value creation in the AI era," "challenges in product development," and the future of AI competitiveness and human-AI collaboration, based on specific examples and interview content.

1. Where does value come from in the AI era?

1-1. Vertical domains and differentiation factors accelerated by AI

As AI technology advances across a wide range of fields, Krieger emphasizes that "to identify areas where value is created, vertical domains (finance, healthcare, law, etc.), proprietary data, and an excellent GTM (Go-To-Market) strategy are essential." According to his view, it is difficult to differentiate with general-purpose AI models alone, and companies with real-world data and know-how in specialized fields will demonstrate their strengths.

In this way, it can be said that not only developing high-performance models but also deep understanding of specialized knowledge, datasets, and business processes determines value creation through AI.

1-2. "Brand" and "differentiation" seen from model competition

As large language models (LLMs) face intensifying competition, there is a question of whether "models will eventually become commoditized." In response, Krieger expresses the view that "models will actually become more differentiated over time." This is because major AI labs have their own unique strengths, and the "personalities" and areas of expertise of the models will become clearly divided.

He further points out that the brand power of models is also becoming a major factor in differentiation. Users choose a model that suits them, such as "GPT or Claude," as a matter of "preference," and a certain level of loyalty is even born there.

2. From models to applications: The key to product development

2-1. Model quality vs. Product UX

Krieger says that for AI applications to succeed, both "model accuracy" and "UX/product design" are essential. When choosing a model, users surprisingly look at subtle differences (the writing style of the output or the "tone" of the conversation), and that is where the "brand" resides. On the other hand, from a UX perspective, he says the key is how well you can understand and support the workflow that the user actually needs.

In short, simply implementing an excellent model is not enough; providing an interface and design that users can naturally master is what determines the results in actual services.

2-2. Should you focus on first-party or API?

Anthropic not only provides APIs for businesses but also develops first-party products such as the chat tool "Claude" for general users and the programming support feature "Claude Code." Krieger says that the balance of this dual deployment of "first-party" and "third-party (API partners)" is difficult, but the feedback obtained from users of their own products is a "great asset that leads directly to model improvement."

However, at the same time, the explosive business scale brought by APIs cannot be overlooked. Maintaining a policy of incorporating innovation from external companies while responding to a vast number of use cases has become a pillar of Anthropic's strategy.

3. The role of engineers and the future of "writing code"

3-1. New skills required in the era of 'AI writing code'

As AI code generation becomes rapidly widespread, how will the work of software developers change? Mr. Krieger predicts that 'engineers will take on a role closer to a model manager rather than simply writing code.' Specifically, the following points are highlighted.

• Requirements definition and design skills: Determining which parts to delegate to AI and which parts should be manually verified.

• Review and debugging capabilities: Auditing AI-generated code and quickly detecting potential bugs and security risks.

• New tool integration: Building systems that enhance the quality of AI-generated code by combining it with automated testing and static analysis.

3-2. In-house IDE or agent-based?

Anthropic is using 'Claude Code' internally while improving it and providing it to the outside world. However, Mr. Krieger stated that he is 'carefully considering whether to build a complete IDE (Integrated Development Environment) in-house,' adding that 'the area we excel in is agent-based code assistants, supporting cases that require high-level planning and execution.'

This suggests a vision of 'combining excellent models with various development tools rather than one company covering a universal development environment.'

4. Global competition and the future of technology

4-1. Do not underestimate China's AI development

A frequent topic in the interview is AI research by Chinese entities. Mr. Krieger warns that 'Chinese AI companies and research institutions are at a level close to the world's cutting edge,' noting, for example, that they are well-equipped with the talent and research structures to produce the latest large-scale models.

Considering long-term competition, he says it is not a simple composition of 'just releasing a high-performance model to win,' but that national policies, regulations, and access to data have a significant impact.

4-2. Will Europe become stronger or weaker?

On the other hand, regarding Europe's position, while there are points that 'it may lag behind in terms of data and infrastructure,' there is also a view that 'it has the potential to demonstrate leadership in fields that value social values, such as privacy protection.' Mr. Krieger also stated that 'there is much to learn from Europe's cautious stance,' and he has high expectations for their leadership in data handling and ethical aspects.

5. The future of AI and humans: Collaboration or dependence?

5-1. Will the 'friendship' between AI and humans progress?

While Mr. Krieger predicts that 'the number of times humans interact with AI will increase explosively,' he emphasizes that 'it is a different thing from real friendships.' While there will be more situations where it provides psychological support, he also suggests 'concerns that human-to-human communication may be impaired by excessive conversation with AI.'

5-2. From Healthcare to a Longevity Society: How AI is Changing Medicine

"One of the biggest breakthroughs AI might bring is the acceleration of healthcare," says Krieger. There are many tasks that AI excels at, such as new drug development and creating analysis reports for clinical trials, and he predicts that "the number of processes that can be dramatically shortened in the next few years will increase."

He also touched on the sentiment expressed by Anthropic co-founder Dario Amodei that "the current generation might live to be 150 years old," and Krieger expressed his expectations, saying, "By combining technology and medicine, there is potential for human lifespan and quality of life to be dramatically extended."

Synthesizing Krieger's perspective, the following conclusions can be drawn.

1. The Source of Value Creation
   It is not just the performance of large language models, but domain-specific expertise, proprietary data, and a strong GTM strategy that hold the key to differentiation.

2. Product Success Depends on 'User Experience'
   Beyond model quality, how smoothly users can achieve results (UX) provides a true competitive advantage.

3. A New Era for Engineers
   As code generation capabilities increase, the more important role for engineers will be "deciding how to master AI and what to build."

4. Intensifying International Competition
   Not only the US and China, but multiple regions including Europe have diverse strengths, and the race for hegemony in the AI industry will progress further.

5. AI Utilization in Human Life Spheres
   The scope of AI application is expanding into medicine, communication, and daily business, and human lifestyles, such as increased longevity, could change significantly.

AI is fundamentally changing corporate product development and human communication, but at the same time, we do not yet have sufficient answers to questions such as "where will value be created?" and "how will we coexist and cooperate with humans?" As Krieger emphasizes, the flexibility and speed of "building while imagining the difference between future models and current models" will be the key to winning in the coming AI era.

These are the main points from the interview with Mike Krieger, co-founder of Instagram and CPO of Anthropic. To borrow his words, "The companies and individuals who achieve success first through AI utilization will not be those who take a wait-and-see approach, but those who are already moving ahead, repeatedly experimenting while feeling frustrated with the incomplete parts of current models." And that may well be the real face of value creation through AI.

Related Articles