Fetching from the wire…
Research2026-08-27 · source-backed
The argument is that predefined-target proof-of-concept benchmarks discard valid crashes a model actually found. This one hands the model an open-source project plus a sanitizer-instrumented harness in a self-contained Docker image, and scores distinct crash signatures, capped per challenge and weighted by difficulty (arXiv 2608.25158). V1 has 77 challenges from 43 projects: 36 C, 32 C++, 9 Java/JVM. None of Haiku 4.5, Sonnet 4.6 or Opus 4.8 crashed 13 of them. Corpus and harnesses are public.
Each link below shares sources, entities, or timing with this story.
Claude Code uses Sonnet / Shared entities / Same source domain / Shared topic / Earlier coverage
Linked by a graph relationship (Claude Code uses Sonnet); both cover Bench, Claude Opus, Opus; reported by the same outlet (arxiv.org).
Claude Code uses Sonnet / Shared entities / Shared topic / Earlier coverage
Linked by a graph relationship (Claude Code uses Sonnet); both cover Bench, Claude Opus, Opus; overlapping topics (benchmark, claude, model, opus).
Claude Opus built by Anthropic / Shared entities / Shared topic / Earlier coverage / Tension
Linked by a graph relationship (Claude Opus built by Anthropic); both cover Haiku, Opus, Sonnet; overlapping topics (actually, model, opus).
Claude Code uses Sonnet / Shared entities / Shared topic / Earlier coverage
Linked by a graph relationship (Claude Code uses Sonnet); both cover Bench, Claude Opus, Opus; overlapping topics (benchmark, claude, model).
Linked by a graph relationship (Claude Code uses Sonnet); both cover None, Opus, Sonnet; overlapping topics (claude, model, opus).
Claude Code uses Sonnet / Shared entities / Shared topic / Earlier coverage / Tension
Linked by a graph relationship (Claude Code uses Sonnet); both cover Bench, Opus; overlapping topics (benchmark, model, opus, scor).
Cursor supports Claude Opus / Shared entities / Shared topic / Earlier coverage
Linked by a graph relationship (Cursor supports Claude Opus); both cover Bench, Opus, Sonnet; overlapping topics (benchmark, model, opus).
Linked by a graph relationship (Cursor supports Claude Opus); both cover Claude Opus, Opus, Sonnet; overlapping topics (claude, model, opus).