Opus 4.5
AI overview
Sign in and the AI will write an overview from our coverage.
Headlines · 2
- Anthropic Outlines an Eight-Step Framework for Evaluating AI Agents
A secondary Juejin AI summary of Anthropic Engineering’s “Demystifying evals for AI agents,” which draws on the company’s internal experience and problems encountered by frontier customers. It covers evaluation tasks, trials, graders, transcripts, outcomes, three grader types, capability and regression evaluations, and metrics for nondeterministic behavior. The eight-step roadmap recommends starting with real failures, using unambiguous tests and clean environments, designing graders carefully, reviewing transcripts, and making evaluations part of team practice. The summary also recommends combining automated evaluations, production monitoring, and human review because each has blind spots.
掘金 · 人工智能 · 🔥 0 - Anthropic and DeepSeek propose compatible agent architectures
A commentary argues that Anthropic’s recommendation to fold subagents into a single agent loop and DeepSeek Harness’s persistent Router/Worker design address different architectural layers rather than contradict each other. Anthropic emphasizes a single loop with skills and tools for capabilities that stronger models can handle internally, while DeepSeek proposes a persistent Router coordinating isolated Workers across concurrent tasks and sessions. The commentary also cites performance figures comparing DeepSeek’s Super-agent design with a single-session subagent approach; these figures come from the cited material and were not independently verified in the supplied evidence.
掘金 · 人工智能 · 🔥 0
Experience and discussion from the community
Share my Opus 4.5 experienceAsk about Opus 4.5
Nobody has shared their experience with Opus 4.5 yet.