在公开处思考技术与 AI · thinking about tech & AI, in the openhttps://dull-bird.github.io/https://dull-bird.github.io/writing/cli-schema-first-contributor/https://dull-bird.github.io/writing/cli-schema-first-contributor/我发现了一个几乎没人知道的开源规范提案 CLI Schema——命令行程序的 OpenAPI。这篇讲三件事：它是什么、我怎么用 AI 参与它的开发、以及和一位低频在线的作者打交道时学到的东西。Fri, 24 Jul 2026 00:00:00 GMTzh开源协作https://dull-bird.github.io/en/writing/cli-schema-first-contributor/https://dull-bird.github.io/en/writing/cli-schema-first-contributor/I stumbled onto CLI Schema, an open spec almost nobody has heard of — an 'OpenAPI for command-line programs.' Three things here: what it is, how I used AI to contribute, and what I learned about working with a maintainer who is rarely online.Fri, 24 Jul 2026 00:00:00 GMTenopen sourcehttps://dull-bird.github.io/writing/uneven-future/https://dull-bird.github.io/writing/uneven-future/Gibson 那句话被引用太多次了。但每次真正用上一个新工具，我都更相信它——差距不在时间，在于谁先伸手。Thu, 18 Jun 2026 00:00:00 GMTzh未来https://dull-bird.github.io/en/writing/uneven-future/https://dull-bird.github.io/en/writing/uneven-future/Gibson's line is over-quoted. But every time I actually pick up a new tool, I believe it more — the gap isn't time, it's who reaches first.Thu, 18 Jun 2026 00:00:00 GMTenthe futurehttps://dull-bird.github.io/writing/slow-bird/https://dull-bird.github.io/writing/slow-bird/我不是最快的那只，但我可以是先起飞的那只。公开地笨拙，比私下地完美更有用。Sat, 30 May 2026 00:00:00 GMTzh方法https://dull-bird.github.io/en/writing/slow-bird/https://dull-bird.github.io/en/writing/slow-bird/I'm not the fastest bird, but I can be the one that starts early. Publicly clumsy beats privately perfect.Sat, 30 May 2026 00:00:00 GMTenmethodhttps://dull-bird.github.io/writing/agent-friendly-web/https://dull-bird.github.io/writing/agent-friendly-web/如果一个 AI agent 来读我的网站，它能顺利拿到它要的东西吗？我把这当成一条新的可访问性标准。Mon, 11 May 2026 00:00:00 GMTzhAI agentshttps://dull-bird.github.io/en/writing/agent-friendly-web/https://dull-bird.github.io/en/writing/agent-friendly-web/If an AI agent visits my site, can it get what it needs? I treat this as a new axis of accessibility.Mon, 11 May 2026 00:00:00 GMTenAI agentshttps://dull-bird.github.io/writing/tools-2026/https://dull-bird.github.io/writing/tools-2026/一份会过时的清单。正因为会过时，才值得记录——它是我此刻的切片。Thu, 02 Apr 2026 00:00:00 GMTzh工具https://dull-bird.github.io/en/writing/tools-2026/https://dull-bird.github.io/en/writing/tools-2026/A list that will age. Which is exactly why it's worth keeping — a slice of right now.Thu, 02 Apr 2026 00:00:00 GMTentoolshttps://dull-bird.github.io/papers/deep-learning-enhanced-index-tracking/https://dull-bird.github.io/papers/deep-learning-enhanced-index-tracking/这篇论文把增强指数跟踪写成一个动态再平衡问题：先按市值选出指数成分股，再让一个结构化神经网络在跟踪误差、超额收益、CVaR 风险和交易成本之间做权重分配。Fri, 07 Jun 2024 00:00:00 GMTzhdeep learningindex trackingportfolio selectionquantitative financehttps://dull-bird.github.io/en/papers/deep-learning-enhanced-index-tracking/https://dull-bird.github.io/en/papers/deep-learning-enhanced-index-tracking/This paper formulates enhanced index tracking as a dynamic rebalancing problem: select large index constituents, then let a structured neural network allocate weights under tracking-error, excess-return, CVaR, and transaction-cost considerations.Fri, 07 Jun 2024 00:00:00 GMTendeep learningindex trackingportfolio selectionquantitative financehttps://dull-bird.github.io/papers/deep-learning-quadratic-hedging/https://dull-bird.github.io/papers/deep-learning-quadratic-hedging/这篇论文不是发明一个新对冲模型，而是系统比较 Multi-net、Single-net 和 RNN 三种深度学习架构在动态对冲里的表现，回答高维、长期限场景下该选哪种网络。Thu, 01 Sep 2022 00:00:00 GMTzhdeep learningquadratic hedgingderivativesRNNbenchmarkhttps://dull-bird.github.io/en/papers/deep-learning-quadratic-hedging/https://dull-bird.github.io/en/papers/deep-learning-quadratic-hedging/This is a benchmark paper rather than a new hedging trick: it compares Multi-net, Single-net, and RNN architectures for dynamic hedging across high-dimensional and long-horizon settings.Thu, 01 Sep 2022 00:00:00 GMTendeep learningquadratic hedgingderivativesRNNbenchmark