AgentEnv, Scale AI's framework for testing agents on tasks and scoring the results, now supports Vercel Sandbox from version 0.9.1334. The provider runs the agent and its task environment in separate sandboxes and records the agent's actions for review.
DigitalOcean has opened a public preview of Learning Services and Agent Data Services. The tools let developers build, evaluate and keep improving AI agents on a single cloud platform.
On the Latent Space podcast, Supabase co-founders Paul Copplestone and Ant Wilson discuss how AI agents producing more software changes what a backend must do. Topics include running long-lived workloads next to the database and making schemas easier for agents to understand.
Luxury resale site The RealReal is tightening its bot defences, making it harder for AI agents to shop there. The change follows a Verge report showing an agent could exploit the retailer's item reserve system.
Entire, the start-up founded by former GitHub CEO Thomas Dohmke, has launched its alternative to GitHub. The platform is built around agents writing code and rethinks the traditional pull request.
Nace AI has applied its Drex decision models to document search for AI agents. The company reports a 92.05% score on BrowseComp-Plus using 12.6 search calls on average, and fewer agent turns on OfficeQA Pro.