"把 Tokenizer 视为 checkpoint ABI，用可验证的迁移契约原地扩展预训练模型词表，并守住生成质量与真实性能。"
开放权重不是单纯公开模型，而是一套产业分发策略。本文用采用、衍生生态、部署成本、硬件覆盖和商业转化五个指标判断它是否有效。
China's open-weight strategy is a distribution system. Evaluate it through adoption, derivatives, deployment cost, hardware reach, and business conver
"A practical contract for expanding a pretrained LLM tokenizer in place while preserving token identity, model quality, and real inference performance
用内容、发现、调用与运行时四层模型，判断 SKILL.md 在 Codex、Claude Code、Gemini CLI 和 GitHub Copilot 之间到底能移植什么。
A four-layer compatibility model for moving SKILL.md workflows across Codex, Claude Code, Gemini CLI, and GitHub Copilot without confusing shared synt
AI Coding Agent 可能在代码审查前执行依赖。本文给出验证包名、来源、版本、哈希和安装脚本的 fail-closed 准入方案。
AI coding agents may execute dependencies before review. Build a fail-closed pre-install gate for package identity, source, version, hashes, and scrip
从 Singular Bank 的客户案例看金融 AI：速度只是第一层价值，批准数据源、结构化工作流、输出留痕与人类判断共同构成可验证闭环。
How Singular Bank reports saving bankers 60–90 minutes daily with ChatGPT and Codex, and why verifiable workflows matter in finance.
