By playing, you agree to uListen's Terms and Privacy Policy and YouTube's Terms of Service.
Sequoia CapitalAnthropic's Katelyn Lesse & Angela Jiang: Building an Ecosystem, not a Walled Garden
Summaries, chapters, transcripts, and FAQs on this page are made by uListen, not YouTube.
At a glance
WHAT IT’S REALLY ABOUT
Anthropic’s platform roadmap: modular primitives evolving from execution to coordination strategies
- Anthropic’s Platform team supports both external developers via APIs and internal product teams via shared infrastructure and primitives.
- The speakers describe a three-layer abstraction stack—knowledge, execution, and coordination—and expect future platform work to move increasingly into coordination via composable “strategies.”
- They argue form factors are rapidly changing (from chat to agents and beyond), so the platform should prioritize robust primitives, standards, and experimentation rather than a single fixed product interface.
- They position Claude Managed Agents and related tooling as packaging of infrastructure and harness best practices to make long-running, agentic execution easier and more reliable.
- They emphasize building an open, interoperable ecosystem through modular infrastructure choices, specs like MCP, connectors, and evolving standards including collaboration on safety-related interoperability.
IDEAS WORTH REMEMBERING
5 ideasAnthropic frames its platform roadmap as a shift from knowledge → execution → coordination abstractions.
Angela Jiang describes a three-layer “cake” of abstractions: knowledge (understanding and shaping interactions with Claude), execution (getting work done with infrastructure and harnesses), and coordination (orchestrating multiple roles/jobs across tokens via higher-level “strategies”). She suggests Anthropic’s roadmap will increasingly move up this stack over time.
The platform is designed for both internal velocity and broad external builder enablement.
Angela Jiang says the team has “two North Stars” because they serve both internal and external builders: internally, maximizing leverage and speed for shipping products; externally, giving any builder tools to work with Claude and build what they want. She links the external focus to being close to where businesses operate, including close integrations with hyperscalers, and to providing primitives and standards.
They try to externalize the same primitives they use internally, validating with both audiences at once.
Both speakers emphasize keeping internal and external primitives aligned so the platform doesn’t overfit to either internal requirements or a single market form factor. Katelyn Lesse adds they often dogfood internally while simultaneously running early access externally to gather a wider range of feedback.
The shift from “messages” to “agents” is driven by repeatable infrastructure and harness needs across customers.
Katelyn Lesse explains the platform evolved from a mostly stateless messages API toward higher-order abstractions aimed at agentic, long-running work. She highlights infrastructure (secure sandboxes, session storage/resume) and harness engineering (prompt caching, context management, cost management) as recurring problems they can package for developers.
Anthropic aims for modular, plug-in architecture rather than a walled garden for execution infrastructure.
Katelyn Lesse says they aren’t “precious” about running everything on Anthropic infrastructure; she cites self-hosted sandboxes and partnerships that let developers plug in other providers, plus capabilities like calling MCP servers behind a firewall. What matters more, she says, is the agent architecture and reliable/scalable interfaces.
WORDS WORTH SAVING
5 quotesExternally, you know, we really are oriented around just helping you just be able to build, but internally, that orientation, while still existing, is probably more, you know, specified towards speed and being able to move really quickly.
— Angela Jiang
We found a lot of our customers solving the same problems over and over again that we also were solving over and over again around as the models got better at running for longer and working with more context at a given time, you want to build agents that can succeed in a kind of long-running context and even a remote context that doesn't necessarily have a human in the loop.
— Katelyn Lesse
So you have knowledge, and you have execution, and you have coordination.
— Angela Jiang
We actually aren't precious about, you should run these things on our infrastructure.
— Katelyn Lesse
The important part is all the kind of context engineering and, like, architecture that we put underneath the hood so that Tag just works.
— Angela Jiang
SPEAKERS
Angela Jiang
guestAnthropic Platform leader focused on developer-facing primitives, abstractions, and ecosystem strategy for Claude.
Sonya Huang
hostSequoia Capital partner who hosts interviews with AI builders and leaders.
Katelyn Lesse
guestAnthropic Platform leader working on Claude’s APIs, developer platform, and internal product infrastructure.
Lauren Reeder
hostSequoia Capital partner and co-host who interviews founders and platform builders.
High quality AI-generated summary created from speaker-labeled transcript.
