Opus 4.8

Headlines · 2

1. 23h
   Developers are uninstalling Superpowers, the Claude Code skill plugin

   A Juejin column argues that Superpowers, described as the most popular skill plugin in the Claude Code ecosystem, is being uninstalled by developers, including the author himself. The plugin injects 14 skill modules that force every task through brainstorming, spec writing, planning, TDD, coding and review, which the author says makes even trivial edits take far longer; the piece cites a user test in which startup alone consumed about 22,000 tokens, roughly 11% of the context window. The author contends that newer models such as Fable 5 now plan and verify their own work, so skills should be enabled on demand rather than installed as a full suite, and notes that the maintainers cut the 14 skills from 3,150 lines of code to 977. Claims such as 250,000 GitHub stars and a collective uninstall are the author's assertions and were not independently verified.

   掘金 · 人工智能 · 🔥 3
2. 3d
   Redwood Research unveils Screwtape, a pipeline for red-teaming AI monitors

   Redwood Research has introduced Screwtape, a pipeline that inserts an attack into a real agent transcript and repeatedly refines it until an action-only monitor no longer catches it. An attacker agent edits the transcript, while an LLM verifier with the attack description and edit diff checks whether the attack is realistic, clearly harmful and plausibly successful; accepted edits are then scored by the monitor. Using monitors based on Opus 4.8 and Gemini 3.1 Pro, Redwood ran the pipeline 252 times on 94 unique transcripts, producing 35 attacks that evaded its own monitors, six of which also evaded a held-out monitor from a frontier developer and were confirmed through human review; Redwood is not releasing the implementation for now but says it is open to sharing it with researchers.

   Redwood Research · 🔥 0

Experience and discussion from the community

Share my Opus 4.8 experienceAsk about Opus 4.8