SC
Anthropic CPO Mike Krieger: Building AI Products From the Bottom Up
@ Sequoia Capital
05/09/25
Takeaways
MK
Mike Krieger
05/09/25
@ Sequoia Capital
Regardless of the medium, the key to AI content is whether there's a compelling story and a relatable person behind it.
MK
Mike Krieger
05/09/25
@ Sequoia Capital
Most content will be AI-generated, making the distinction of whether something was made by AI less useful in the future.
MK
Mike Krieger
05/09/25
@ Sequoia Capital
In product development, solving real problems for users remains crucial, whether for developer tools or end-user products.
MK
Mike Krieger
05/09/25
@ Sequoia Capital
At Anthropic, we embrace a bottoms-up approach to creativity, allowing for more innovative product development close to the model's capabilities.
MK
Mike Krieger
05/09/25
@ Sequoia Capital
The MCP product emerged from engineers recognizing the need for a standardized protocol to better interact with models, rather than a top-down directive.
MK
Mike Krieger
05/09/25
@ Sequoia Capital
The future of AI products will focus on enabling actions and automating workflows, not just retrieving information.
MK
Mike Krieger
05/09/25
@ Sequoia Capital
As coding agents improve, they highlight inefficiencies in product organizations, making alignment meetings more critical and time-consuming.
MK
Mike Krieger
05/09/25
@ Sequoia Capital
Internal products that share visibility, like those integrated with Slack, encourage the use of AI tools without stigma, fostering a culture of collaboration.
MK
Mike Krieger
05/09/25
@ Sequoia Capital
We're still at the very beginning of how people understand how to use AI tools in their work, and there are many social opportunities that haven't been fully explored yet.
MK
Mike Krieger
05/09/25
@ Sequoia Capital
Our goal is to enable models to work autonomously for longer periods, which will require advancements in memory, tool use, and organizational onboarding.
MK
Mike Krieger
05/09/25
@ Sequoia Capital
The importance of compute allocation is a critical topic at Anthropic, as it influences our strategic decisions on whether to focus on research or product development.
MK
Mike Krieger
05/09/25
@ Sequoia Capital
There's a challenge in balancing the release of models for external use versus deploying them internally for greater utility.
MK
Mike Krieger
05/09/25
@ Sequoia Capital
I believe that having our models out in the wild has provided valuable market feedback, which has shaped our product development significantly.
MK
Mike Krieger
05/09/25
@ Sequoia Capital
Many users are comfortable paying for multiple AI services, and while this may be sustainable in the short term, there could be a future desire for consolidation.
MK
Mike Krieger
05/09/25
@ Sequoia Capital
Products that start as AI-light and transition to AI-heavy often struggle to integrate AI features seamlessly into their core functionality.
MK
Mike Krieger
05/09/25
@ Sequoia Capital
A surprising number of AI-native products fail to expose the necessary application primitives to the models, limiting their effectiveness.