What matters in AI.

Subscribe

Research

Claude Opus 4.8 follows 70.1% of coding rules in a new test

In this test, agents most frequently do not follow the rules for the sequence of work, tools, and reports.

Claimed, not confirmed

Best coding agent passes every check on 19.0% of tasksClaude Opus 4.8 under Cline, on 358 tasks, passes every constraint check on 19.0% of tasks. A grid of 100 squares shows 19 filled. Two bars show it meets 70.1% of single constraints and 63.7% of relation constraints. Claude Opus 4.8 under Cline · 358 tasks 19.0% of tasks pass every check Atomic compliance 70.1% Relation compliance 63.7%
On the ANCHOR benchmark, the best agent, Claude Opus 4.8 under Cline, passes every constraint check on just 19.0% of 358 tasks.

Sources

  1. ANCHOR: Assessing Naturalistic Constraint Hierarchies and Ordering in Repository-Grounded Agentic Codingarxiv.org
AI MATTER · NEWS · AI MATTER · NEWS ·14 OCT2026

Posted

Tags

More in Research

All Research news