Show HN: Kronumos-Neurosymbolic code repair with a 5µs Rust Sub-Cortex

AI Digest

Kronumos 2 Kairos通过结合代码模型与子皮质层技术，在SWE-bench验证数据集上实现低成本高效代码修复，修复成功率超80%。

Kronumos 2 Kairos combines code models with sub-cortex tech to achieve cost-effective automated repair on SWE-bench, resolving 8+ production bugs with 0.00$ cost.

Key points

• 融合7B参数代码模型与零分配子皮质层架构 Integrates 7B code model with zero-allocation sub-cortex architecture
• 在SWE-bench验证数据集实现442个修复候选 Generates 442 repair candidates on SWE-bench dataset
• 通过AST修复器提升33%修复成功率 AST healer boosts resolution rate by 33%
• 单任务平均耗时2.5秒且成本趋近零 Single-task cost approaches $0 with 2.5s latency

Takeaway: 低成本高效率的代码修复方案，适合大规模软件维护场景 / Cost-effective repair solution with 80%+ success rate for software maintenance

Why it matters 展示如何通过混合架构实现工业级代码修复，提供可复用的技术框架