Software 1.0 是手写代码;Software 2.0 是神经网络的权重;Software 3.0 是 prompt。
Software 3.0, May 2026
我的看法这个分类很顺,但它低估了转换成本。把 SOP 翻译成 prompt 比把伪代码翻译成代码难得多——它要捕捉的是隐性知识,不是显性逻辑。
§ 01
News
AI 圈每天发生了什么2026.11.14arXiv cs.AI
Zugwise: Domain-Aware Game Reconstruction for Digitizing Handwritten Chess ScoresheetsDigitizing handwritten chess scoresheets has been studied as a handwriting-recognition problem: read the move in each box and check that it is legal. This framing misses the errors that matter most. A misread that yields an illegal move is loud. One that yields a legal-but-wrong move is silent, because a legality-only…
↗2026.11.14arXiv cs.AICleaning an Agent's History Hurts When the World Still Holds the ErrorTool-using language-model agents accumulate histories of their own mistakes, and a system that compacts, repairs or hands off such a history decides what it says about those mistakes while their effects stay in the world. Prior work cleans the history at a fixed world, where showing a corrected step and saying it is d…
↗2026.11.14arXiv cs.AITogether Compiler: Continual Compilation for AI InferenceToday's AI workloads are rapidly changing: new models are released more often, model architectures are increasingly diverse, and traffic distributions change frequently. The challenge we face at Together AI is continual optimization for these changing workloads. Compiling a deployment into an optimized forward pass is…
↗2026.11.14arXiv cs.AIWhen Should Semantic Priors Enter Partially-Identified Causal Discovery?Observational causal discovery often leaves causal di- rections unresolved or unstable. Large language mod- els (LLMs) can provide semantic information to resolve these directions, but incorrect proposals can exclude the true causal graph and invalidate downstream inference. We study three separate decisions: where to…
↗2026.11.14arXiv cs.AISynthCorp: A Platform for Creating Benchmarks for Data AgentsCurrent benchmarks for LLM-driven data agents saturate quickly and cannot target specific capabilities. We present SynthCorp, a platform that creates customizable data estates and benchmarks by simulating how business events produce and modify corporate data over time. Each data estate combines relational databases, a…
↗2026.11.14arXiv cs.AIRational Use of Time for Social Coordination in Humans and AIAll social interactions are situated in time. Previous studies have shown that humans are able to use implicit temporal cues, such as the duration of a pause, to infer others' mental states. But how do humans use temporal cues for dynamic, fine-grained social coordination? And can AI systems do the same, without ``exp…
↗2026.11.14arXiv cs.AILeaderboard Hacking: Preference-Based Model Evaluations are Vulnerable to ManipulationArena-style rankings of language models are a widely used evaluation framework in online leaderboards, research papers, and model evaluation in production. These rankings aim to realistically assess model performance using pairwise comparisons of anonymous models on user prompts, aggregating votes via the Bradley-Terr…
↗2026.11.14arXiv cs.AIRobust Task-Aware State Estimation from Pre-trained PerceptionAgents that perform state estimation using pretrained perception models, e.g., Vision-Language Models (VLMs), face the fundamental challenge that predictions are uncertain. Still, most approaches determinize these predictions into a single best-guess state and make decisions as if it were ground truth. This is brittle…
↗2026.11.13Hacker News (AI)No AI Slop: A Manifesto
↗2026.11.13Hugging FaceBuild a Robot Pick-and-Place Simulation with SimReady Assets
↗§ 03
Perspectives
他者之声这个时代最持久的能力,是学习任何你想学的东西。
Almanack, 2020
我的看法十年前对,今天更对。当工具的 half-life 缩短到 18 个月,可迁移的"元能力"才是真正的复利。
§ 04
Plans
正在进行Carol's Website
用 Claude Code 把这个站做出来,两天 MVP。一次从 idea 到 product 的完整练习,你正在看的就是它。
Shipped · 2026
The Big Inventory of AI Products
2025 年底做的一份 AI 产品分类图谱,被几个朋友长期收藏。
Shipped · 2025
AI 产品周记
每周一份精选短评,记录我在 AI 产品圈看到的、想到的、被反复触发的判断。
In Progress
数字分身 Agent v2
从全文 prompt 升级到 RAG,加上对话历史。让分身真的能"代我回答"。
Exploring
§ 05
About
关于这份日报清华本科 · 大厂 AI PM · 亿级用户产品增长 · AI Pilled
我是 Carol,白天做 AI 产品,晚上把看见的、想到的、喜欢的人和观点写下来。这个站不是博客,更像一份给自己也开放给别人的 AI-Pilled 日报。
AI 让信息变多、让判断变少。所以这份日报每天替你过一遍,然后把我看完之后真正想说的那部分留下来。每一篇都希望值得被读两遍.
如果你也在思考 AI 与产品、与表达、与生活方式的关系,欢迎留个联系方式,或者直接去跟我的数字分身聊聊。