Simon Willison quotes Addy Osmani, Morgan Westlee Lunt and Qing Wang on how the Claude Code team approaches code review. Rather than asking whether a human read the code, they ask which check, automated or human, best catches what could go wrong, so most changes are reviewed and fixed by Claude before a person looks at them.
Replit CEO Masad said the company has stopped working with Anthropic, citing several reasons including cost. Replit announced a partnership with OpenAI in August, but nothing had suggested the Anthropic relationship had ended.
This SiliconANGLE analysis argues that security and data resilience, long handled separately, are merging because of agentic AI. The author says most organizations have not yet noticed the shift.
Perplexity has released SPACE, the sandbox that runs every task in Perplexity Computer, under the Apache 2.0 license. Each agent task gets its own Firecracker microVM, with credentials and outbound traffic controlled from outside it, which could help teams building coding or browsing agents.
Google researchers have open-sourced EnvHarness, a framework that adjusts agent-training environments to an agent's weak spots without changing the agent, the environment or the evaluation setup. The researchers say it can make learning both more effective and more efficient.
Austin-based RapidCanvas, which combines an AI agent platform with its own engineers to build business apps, raised a $29M Series B led by Vertex Ventures SEA & India, with Hitachi Ventures taking part. It plans to hire in the US, Brazil and India, bringing total funding to $52.5M.