Announcement

• Sep 01, 202614 min read

  Introducing FrontierHarness Eval

  FrontierHarness compares 9 production coding harnesses across 12 configurations on 30 agentic tasks, holding the model, tasks, and runtime constant. Pass rates cluster in a 17-point band. What separates the harnesses is how many tokens and steps they spend to get there: 17x from cheapest to most expensive per completed task.

  agentsannouncementbenchmarkharnessevaluation
• Aug 13, 20261 min read

  Introducing SSH access for Runta runtimes

  Connect any SSH-capable agent to a persistent Runta runtime, with access to services but without exposing the underlying credentials.

  agentsannouncementruntimesecurityssh
• Jul 16, 20264 min read

  Runta: the execution layer for agents

  Runta is your execution layer for AI agents: improving token efficiency, protecting secrets, and governing access. Today, we're announcing $20M in seed funding led by Martin Casado at Andreessen Horowitz.

  agentsannouncementexecution layer