After the cost per run of its AI code reviewer doubled, Cloudflare analysed its token usage. Prompt caching and routing requests to the right model cut its inference spending by 68% and made reviews faster.
Eugenia Kuyda, founder of Replika, has launched a new startup, Wabi, built around an idea she had more than a decade ago. Wabi offers a free personal AI assistant that anyone can download now.
GitHub explains how to hand off an agent session from the GitHub Copilot app to VS Code. The conversation, code changes and working context carry over intact.
Cognizant and Google Cloud have set up a Gemini Frontier Capability Group with more than 18,000 Cognizant employees building AI agents for clients on Gemini. The work will cover legacy system modernisation, developer tools, process automation and cyber defence; no financial terms were disclosed.
The web used to split its users into bots and people, with long-standing rules for automated tasks such as page scraping. MIT Technology Review looks at how AI agents blur that line, and how reshaping websites for them could carry costs for human users.
The authors of AI Snake Oil have published a paper on how AI agents are evaluated. It surveys the methods in use, the open questions and the wider ecosystem around agent evaluation.