Hacker NewsLeo de Moura Formal Verification for AI Code via LeanLeo de Moura Blog·high signalLean creator argues formal verification is the only scalable answer to AI code that half fails security tests. 249 pts.SourceSource pageLeo de Moura Blog↳ Follow the threadPolicy dependency / Stack layerReason Less, Verify More: Deterministic Gates Catch a Silent Policy-Violation Failure in Tool-Using AgentsarXivPolicy dependency / Stack layerKastra Launches Policy Enforcement for Claude Code, Cursor, and CodexHacker NewsPolicy dependency / Stack layerRoots, Sampling, and Logging are deprecated — new capabilities now ship as opt-in extensionsModel Context Protocol BlogStack layer / Threat patternEvoAgentBench: A Benchmark That Isolates Whether Agents Can Turn Past Sessions Into Reusable ProceduresarXivStack layer / Threat patternllm 0.31.1 Fixes a Tool Call Crash That Only Shows Up on Non-OpenAI Providerssimonwillison.netPolicy dependency / Stack layerStrict data-flow tracking is the only defense that zeroes out ADI — spend its utility cost only on code execution and moneyarXiv 2607.05120Stack layer / Threat patternMeta Charges for a Frontier Model for the First Time: Muse Spark 1.1 at $1.25/$4.25TechCrunchPolicy dependency / Stack layerPrismata Confines Cross-Site Prompt Injection by Treating Web Agents Like an XSS ProblemarXiv