Policy dependency / Stack layer
OpenAI publicly brakes its own frontier run: two-week RL pause after 'Astra' couldn't be ruled out of the Critical cyber tier
OpenAI
Stack layer / Contrast
Five 2026 GenAI Systems All Beat the Average Student on Intro OOP Exams but Still Break on Interfaces and Abstract Classes
arXiv 2608.16318
Policy dependency / Contrast
ADEPT pre-trains one dexterity policy on object reposing, then post-trains downstream skills onto two robot hands
arXiv
Stack layer / Update thread
Jeremy Morrell Argues LLMs Finally Make User-Written Extensions Economical, and Points at Salesforce Apex From 2007 as Proof
jeremymorrell.dev / Hacker News
Stack layer / Update thread
codebase-memory-mcp v0.10.6 Reverses a Four-Release Indexing Regression: Java 477s → 86s, C# 1211s → 450s
GitHub
Stack layer / Threat pattern
How you lay out your repo changes prompt-injection success: highly modular workspaces measurably lower attack success rate
arXiv 2608.14876
Policy dependency / Stack layer
Claude Code Tightens macOS Sandbox: Wildcard Read-Deny Rules Now Beat Allowed Read Regions
Claude Code Changelog
Policy dependency / Stack layer
Compose agent guardrails as an algebra instead of a rule list: 94.8% of policy-violating events intercepted while keeping 86.9% task completion
arXiv 2608.16402