<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Ash Knowledge Horizon</title><link>https://alwayssearch.ash-yxh-site.workers.dev/</link><description>项目、研究、方法、学习与策展专题的公开更新。</description><language>zh-CN</language><lastBuildDate>Tue, 04 Aug 2026 16:09:00 GMT</lastBuildDate><atom:link href="https://alwayssearch.ash-yxh-site.workers.dev/rss.xml" rel="self" type="application/rss+xml"/><item><title>可信 AI 与规则治理</title><link>https://alwayssearch.ash-yxh-site.workers.dev/topics/governing-ai-rules/</link><guid isPermaLink="true">https://alwayssearch.ash-yxh-site.workers.dev/topics/governing-ai-rules/</guid><description>从合规风控的业务问题出发，理解规则如何被建模、学习、评测和受控改进，并把 AI 放进可追溯的责任结构。</description><pubDate>Tue, 04 Aug 2026 16:09:00 GMT</pubDate><category>主题专题</category></item><item><title>从“能用”到“放心用”：我对可信 AI 产品边界的复盘</title><link>https://alwayssearch.ash-yxh-site.workers.dev/learning/from-ai-usable-to-trustworthy/</link><guid isPermaLink="true">https://alwayssearch.ash-yxh-site.workers.dev/learning/from-ai-usable-to-trustworthy/</guid><description>在梳理政企智能体与规则智能方案后，我逐渐把可信 AI 理解为一套产品责任结构，而不是单一模型指标。</description><pubDate>Tue, 04 Aug 2026 16:08:00 GMT</pubDate><category>学习与成长</category><category>可信 AI</category><category>AI 治理</category><category>政企 AI</category><category>产品设计</category></item><item><title>规则评测与发布门禁：让每次变更都能被回放</title><link>https://alwayssearch.ash-yxh-site.workers.dev/methods/rule-evaluation-and-release-gates/</link><guid isPermaLink="true">https://alwayssearch.ash-yxh-site.workers.dev/methods/rule-evaluation-and-release-gates/</guid><description>建立固定数据集、分群指标、新旧版本差异、审批与回滚记录，为规则学习和规则进化提供共同发布底座。</description><pubDate>Tue, 04 Aug 2026 16:07:00 GMT</pubDate><category>方法与工具</category><category>评测与发布</category><category>规则引擎</category><category>AI 治理</category><category>规则学习</category><category>规则进化</category></item><item><title>从人工判例到规则进化：如何避免在线自学习失控</title><link>https://alwayssearch.ash-yxh-site.workers.dev/methods/rule-evolution-from-cases/</link><guid isPermaLink="true">https://alwayssearch.ash-yxh-site.workers.dev/methods/rule-evolution-from-cases/</guid><description>把人工复核、误报漏报和例外处置转化为受控变更建议，而不是让模型直接修改生产规则。</description><pubDate>Tue, 04 Aug 2026 16:06:00 GMT</pubDate><category>方法与工具</category><category>规则进化</category><category>规则引擎</category><category>AI 治理</category><category>产品设计</category></item><item><title>规则学习不是自动写规则：一条受控生产流水线</title><link>https://alwayssearch.ash-yxh-site.workers.dev/methods/rule-learning-controlled-pipeline/</link><guid isPermaLink="true">https://alwayssearch.ash-yxh-site.workers.dev/methods/rule-learning-controlled-pipeline/</guid><description>从文档解析、条款理解、要素对齐到候选编译和发布，把大模型限制在可验证的规则生产过程内。</description><pubDate>Tue, 04 Aug 2026 16:05:00 GMT</pubDate><category>方法与工具</category><category>规则学习</category><category>规则引擎</category><category>合规建模</category><category>产品设计</category></item><item><title>GRC 平台 PoC 的八组验证问题</title><link>https://alwayssearch.ash-yxh-site.workers.dev/methods/grc-poc-evaluation-framework/</link><guid isPermaLink="true">https://alwayssearch.ash-yxh-site.workers.dev/methods/grc-poc-evaluation-framework/</guid><description>把竞品功能表转化为真实验证脚本，从建模、执行、AI、穿透、闭环、治理、集成和可解释性八个维度开展 PoC。</description><pubDate>Tue, 04 Aug 2026 16:04:00 GMT</pubDate><category>方法与工具</category><category>PoC 验证</category><category>GRC</category><category>产品策略</category><category>评测与发布</category></item><item><title>政企场景的可信 AI：业务语义、可控编排与运行隔离</title><link>https://alwayssearch.ash-yxh-site.workers.dev/research/trusted-ai-in-government-and-enterprise/</link><guid isPermaLink="true">https://alwayssearch.ash-yxh-site.workers.dev/research/trusted-ai-in-government-and-enterprise/</guid><description>依据久其女娲天工 AI 平台公开材料，分析政企智能体从能用走向放心用所需的四层基础。</description><pubDate>Tue, 04 Aug 2026 16:03:00 GMT</pubDate><category>研究与思考</category><category>可信 AI</category><category>政企 AI</category><category>本体与知识</category><category>AI 治理</category></item><item><title>用友 BIP 合规与风险建模：从公共底座到执行闭环</title><link>https://alwayssearch.ash-yxh-site.workers.dev/research/yonyou-bip-compliance-modeling/</link><guid isPermaLink="true">https://alwayssearch.ash-yxh-site.workers.dev/research/yonyou-bip-compliance-modeling/</guid><description>依据公开帮助资料梳理合规项、风险项、智能审核与智能巡检之间的关系，并标出仍需 PoC 验证的边界。</description><pubDate>Tue, 04 Aug 2026 16:02:00 GMT</pubDate><category>研究与思考</category><category>合规建模</category><category>GRC</category><category>规则引擎</category></item><item><title>穿透式监管的三层模型：监管、业务与交易</title><link>https://alwayssearch.ash-yxh-site.workers.dev/research/three-layer-penetrating-supervision/</link><guid isPermaLink="true">https://alwayssearch.ash-yxh-site.workers.dev/research/three-layer-penetrating-supervision/</guid><description>把经常混用的穿透式监管拆成监管层级、业务链路和交易关系三种能力，明确各自的数据与产品前提。</description><pubDate>Tue, 04 Aug 2026 16:01:00 GMT</pubDate><category>研究与思考</category><category>穿透式监管</category><category>GRC</category><category>产品设计</category></item><item><title>企业合规风控平台的四类竞争路径</title><link>https://alwayssearch.ash-yxh-site.workers.dev/research/grc-competitive-landscape/</link><guid isPermaLink="true">https://alwayssearch.ash-yxh-site.workers.dev/research/grc-competitive-landscape/</guid><description>从 ERP 一体化、专业 GRC、监管科技与金融实时风控四条路径理解市场，不把功能清单误当成产品能力。</description><pubDate>Tue, 04 Aug 2026 16:00:00 GMT</pubDate><category>研究与思考</category><category>GRC</category><category>产品策略</category><category>穿透式监管</category></item></channel></rss>
