agents
small, capable agents that take action, adapt, and deliver outcomes.
- applied research
we explore ideas that solve real problems.
- practical agents
we build agents that integrate into work.
- real workflows
we design for how teams actually work.
what we’re building
tools
developer tools and libraries for building agentic systems.
research
technical research that advances practical ai systems.
why subintern
we believe ai should be useful, reliable, and designed for real work.
we build small, capable systems with a focus on useful autonomy, careful evaluation, and thoughtful product design.
learn more about us- useful autonomy
agents that take initiative within clear boundaries and deliver measurable value.
- small, capable systems
we optimize for simplicity, reliability, and compounding capability over scale for its own sake.
- thoughtful product design
interfaces and integrations that feel invisible, so teams can focus on their work.
latest
- projectagent sandbox
a local environment for building, testing, and iterating on agents.
May 12, 2026 - researchworkflow memory
giving agents reliable memory of your workflows without the clutter.
Apr 28, 2026 - articledesigning for useful autonomy
principles and patterns for building agents people can trust.
Apr 14, 2026
products
your agent in a pocket — dump your day and it breaks everything into tasks, then follows through to get them done.
Learn morea context-aware reading layer for the open web. select any term while reading and get a short, context-shaped preview — then save snippets to a synced reading list.
Learn more