diff --git a/README.md b/README.md index e4a981aa27..71c06d77a1 100644 --- a/README.md +++ b/README.md @@ -1,20 +1,21 @@
-

LoopX

+

LoopX 2.0

-**Give your agents a goal. Keep the work moving.** +**The workbench for long-horizon research.** -The open, local-first control plane for long-horizon agents and personal agent teams.
-Keep goals, decisions and evidence across sessions. Work with Codex, Claude Code, DeepSeek Harness and other supported runtimes. +For enterprise teams, research organizations, and independent builders.
+Coordinate agents around complex research goals, preserve decisions and evidence across sessions, inspect outcomes and costs, and carry findings into the next step. Open source and local first, with support for Codex, Claude Code, DeepSeek Harness, and other supported runtimes.
+2.0 development preview · the installation below follows the latest stable release. -[![License](https://img.shields.io/badge/license-Apache--2.0-blue.svg)](LICENSE) [![Release](https://img.shields.io/github/v/release/loopx-project/loopx?filter=v*&display_name=tag)](https://github.com/loopx-project/loopx/releases/latest) [![GitHub contributors](https://img.shields.io/github/contributors-anon/loopx-project/loopx?color=blue)](https://github.com/loopx-project/loopx/graphs/contributors) [![Discord](https://img.shields.io/badge/Discord-Join-5865F2?logo=discord&logoColor=white)](https://discord.gg/XmGgQyCFZd) [![TypeScript core](https://img.shields.io/badge/core-TypeScript-3178C6?logo=typescript&logoColor=white)](docs/architecture/rfcs/typescript-control-plane-migration-v0.md) [![Python](https://img.shields.io/badge/python-3.11%2B-blue.svg)](pyproject.toml) [![Local first](https://img.shields.io/badge/control--plane-local--first-brightgreen.svg)](docs/public-private-boundary.md) [![Loop Agents](https://img.shields.io/badge/status-loop%20agents%20supported-brightgreen.svg)](docs/product/release-readiness.md) +[![License](https://img.shields.io/badge/license-Apache--2.0-blue.svg)](LICENSE) [![Release](https://img.shields.io/badge/releases-GitHub-181717)](https://github.com/loopx-project/loopx/releases/latest) [![GitHub contributors](https://img.shields.io/badge/contributors-GitHub-181717)](https://github.com/loopx-project/loopx/graphs/contributors) [![Discord](https://img.shields.io/badge/Discord-Join-5865F2?logo=discord&logoColor=white)](https://discord.gg/XmGgQyCFZd) [![TypeScript core](https://img.shields.io/badge/core-TypeScript-3178C6?logo=typescript&logoColor=white)](docs/architecture/rfcs/typescript-control-plane-migration-v0.md) [![Python](https://img.shields.io/badge/python-3.11%2B-blue.svg)](pyproject.toml) [![Local first](https://img.shields.io/badge/control--plane-local--first-brightgreen.svg)](docs/public-private-boundary.md) [![Loop Agents](https://img.shields.io/badge/status-loop%20agents%20supported-brightgreen.svg)](docs/product/release-readiness.md) loopx-project/loopx on Trendshift [Get started](#try-loopx) · [Workspace](#meet-the-personal-agent-workspace) · [LHTB results](#lhtb-results) · [Docs](https://loopx-project.github.io/loopx/docs/) · [简体中文](README.zh-CN.md) -LoopX Personal Agent Workspace: a Work map showing task dependencies, responsible agents, progress and decisions waiting for the owner
-Source-built App on an authored Workspace stories replay; no live Agent execution · watch the 32-second walkthrough +LoopX 2.0, the long-horizon research workbench for enterprise teams, research organizations and independent builders; recorded research history in a labelled App frame
+Three candidate films: A · product flow · 79s · B · research story · 83s · C · predict and compare · 87s · transcripts, subtitles and sources. Recorded App excerpts, synthetic examples and illustrations are labelled. **Start in Codex App in two steps.** Install once: @@ -37,8 +38,9 @@ $loopx fix the open PR review feedback and keep the patch reviewable --- -**More verified work. Less human attention.** LoopX gives agents durable goals, -bounded continuation, peer ownership and recoverable handoffs. Your runtime +**More verified work. Less human attention.** A control plane for long-horizon agents +and personal agent teams gives LoopX durable goals, bounded continuation, peer +ownership and recoverable handoffs. Your runtime provides the model and tools; LoopX keeps track of what to do next, what is accepted, and when to ask you. diff --git a/README.zh-CN.md b/README.zh-CN.md index ceaa448e5b..c0dc2e6252 100644 --- a/README.zh-CN.md +++ b/README.zh-CN.md @@ -1,20 +1,21 @@
-

LoopX

+

LoopX 2.0

-**给 Agent 一个目标，让工作持续向前。** +**长程研究工作台** -面向长程 Agent 与个人 Agent 团队的开源、本地优先控制面。
-跨会话保留目标、决策与证据，连接 Codex、Claude Code、DeepSeek Harness 等已有运行时。 +面向企业、研究机构与超级个体。
+围绕复杂研究目标，协调多个 Agent 持续推进，跨会话保留研究过程、决策与证据，检查结果和成本，并将发现用于下一步工作。开源、本地优先，连接 Codex、Claude Code、DeepSeek Harness 等已有运行时。
+2.0 开发预览 · 下方安装命令仍跟随最新稳定版。 -[![License](https://img.shields.io/badge/license-Apache--2.0-blue.svg)](LICENSE) [![Release](https://img.shields.io/github/v/release/loopx-project/loopx?filter=v*&display_name=tag)](https://github.com/loopx-project/loopx/releases/latest) [![GitHub 贡献者](https://img.shields.io/github/contributors-anon/loopx-project/loopx?color=blue)](https://github.com/loopx-project/loopx/graphs/contributors) [![Discord](https://img.shields.io/badge/Discord-Join-5865F2?logo=discord&logoColor=white)](https://discord.gg/XmGgQyCFZd) [![TypeScript core](https://img.shields.io/badge/core-TypeScript-3178C6?logo=typescript&logoColor=white)](docs/architecture/rfcs/typescript-control-plane-migration-v0.zh-CN.md) [![Python](https://img.shields.io/badge/python-3.11%2B-blue.svg)](pyproject.toml) [![Local first](https://img.shields.io/badge/control--plane-local--first-brightgreen.svg)](docs/public-private-boundary.md) [![Loop Agents](https://img.shields.io/badge/status-loop%20agents%20supported-brightgreen.svg)](docs/product/release-readiness.md) +[![License](https://img.shields.io/badge/license-Apache--2.0-blue.svg)](LICENSE) [![Release](https://img.shields.io/badge/releases-GitHub-181717)](https://github.com/loopx-project/loopx/releases/latest) [![GitHub 贡献者](https://img.shields.io/badge/contributors-GitHub-181717)](https://github.com/loopx-project/loopx/graphs/contributors) [![Discord](https://img.shields.io/badge/Discord-Join-5865F2?logo=discord&logoColor=white)](https://discord.gg/XmGgQyCFZd) [![TypeScript core](https://img.shields.io/badge/core-TypeScript-3178C6?logo=typescript&logoColor=white)](docs/architecture/rfcs/typescript-control-plane-migration-v0.zh-CN.md) [![Python](https://img.shields.io/badge/python-3.11%2B-blue.svg)](pyproject.toml) [![Local first](https://img.shields.io/badge/control--plane-local--first-brightgreen.svg)](docs/public-private-boundary.md) [![Loop Agents](https://img.shields.io/badge/status-loop%20agents%20supported-brightgreen.svg)](docs/product/release-readiness.md) loopx-project/loopx 在 Trendshift 的趋势排名 [开始使用](#试用-loopx) · [个人工作区](#认识个人-agent-工作区) · [LHTB 结果](#lhtb-results) · [文档](https://loopx-project.github.io/loopx/docs/) · [English](README.md) -LoopX 个人 Agent 工作区：工作地图展示任务依赖、负责 Agent、进展与等待用户处理的决策
-源码构建的 App 运行预设的 Workspace 场景回放，未执行实时 Agent 任务 · 观看 32 秒完整演示 +LoopX 2.0，面向企业、研究机构与超级个体的长程研究工作台；应用画框内保留真实研究历史与来源范围
+三版候选：A · 产品流程 · 79 秒 · B · 研究故事 · 83 秒 · C · 机制推演 · 87 秒 · 文字稿、字幕与来源。真实 App、合成示例与机制示意分别标注。 **两步在 Codex App 里用上 LoopX。** 先安装一次： diff --git a/docs/assets/personal-workspace/loopx-research-a.en.srt b/docs/assets/personal-workspace/loopx-research-a.en.srt new file mode 100644 index 0000000000..6a5cd3b975 --- /dev/null +++ b/docs/assets/personal-workspace/loopx-research-a.en.srt @@ -0,0 +1,83 @@ +1 +00:00:00,300 --> 00:00:04,089 +Loop X two point zero. A workbench for +long-horizon research. + +2 +00:00:07,300 --> 00:00:10,189 +Carry the goal, constraints, and return +route. + +3 +00:00:10,369 --> 00:00:12,440 +Let the existing owner adopt the work. + +4 +00:00:12,620 --> 00:00:15,602 +Keep corrections on the task. Return the +evidence. + +5 +00:00:18,300 --> 00:00:21,287 +See work, waiting decisions, and schedules. + +6 +00:00:22,800 --> 00:00:25,171 +Separate the next task from a pending +approval. + +7 +00:00:26,800 --> 00:00:28,667 +Keep owners and waiting work visible. + +8 +00:00:30,300 --> 00:00:33,723 +Keep questions, findings, and changes of +direction. + +9 +00:00:35,800 --> 00:00:39,046 +Read six process judgments. Keep unknowns +visible. + +10 +00:00:39,226 --> 00:00:42,872 +Trace claims to their source. Measure +results and costs separately. + +11 +00:00:45,300 --> 00:00:49,877 +One generated-graph probe improved from +forty one point two to eighty five percent. + +12 +00:00:50,057 --> 00:00:53,062 +That result applies to the duplication +condition shown. + +13 +00:00:53,700 --> 00:00:58,099 +The independent control kept its accuracy. +Time rose thirty two percent. + +14 +00:00:58,279 --> 00:01:00,236 +Native score gain remains unproven. + +15 +00:01:00,900 --> 00:01:04,091 +Keep failed paths and their reasons. Choose +the next test. + +16 +00:01:06,400 --> 00:01:09,359 +Check continuation, hosts, and channels. + +17 +00:01:10,000 --> 00:01:12,163 +Your runtime brings the models and tools. + +18 +00:01:14,000 --> 00:01:18,033 +Loop X. Keep the research, evidence, and +next decision together. diff --git a/docs/assets/personal-workspace/loopx-research-a.en.vtt b/docs/assets/personal-workspace/loopx-research-a.en.vtt new file mode 100644 index 0000000000..5adea0b897 --- /dev/null +++ b/docs/assets/personal-workspace/loopx-research-a.en.vtt @@ -0,0 +1,85 @@ +WEBVTT + +1 +00:00:00.300 --> 00:00:04.089 +Loop X two point zero. A workbench for +long-horizon research. + +2 +00:00:07.300 --> 00:00:10.189 +Carry the goal, constraints, and return +route. + +3 +00:00:10.369 --> 00:00:12.440 +Let the existing owner adopt the work. + +4 +00:00:12.620 --> 00:00:15.602 +Keep corrections on the task. Return the +evidence. + +5 +00:00:18.300 --> 00:00:21.287 +See work, waiting decisions, and schedules. + +6 +00:00:22.800 --> 00:00:25.171 +Separate the next task from a pending +approval. + +7 +00:00:26.800 --> 00:00:28.667 +Keep owners and waiting work visible. + +8 +00:00:30.300 --> 00:00:33.723 +Keep questions, findings, and changes of +direction. + +9 +00:00:35.800 --> 00:00:39.046 +Read six process judgments. Keep unknowns +visible. + +10 +00:00:39.226 --> 00:00:42.872 +Trace claims to their source. Measure +results and costs separately. + +11 +00:00:45.300 --> 00:00:49.877 +One generated-graph probe improved from +forty one point two to eighty five percent. + +12 +00:00:50.057 --> 00:00:53.062 +That result applies to the duplication +condition shown. + +13 +00:00:53.700 --> 00:00:58.099 +The independent control kept its accuracy. +Time rose thirty two percent. + +14 +00:00:58.279 --> 00:01:00.236 +Native score gain remains unproven. + +15 +00:01:00.900 --> 00:01:04.091 +Keep failed paths and their reasons. Choose +the next test. + +16 +00:01:06.400 --> 00:01:09.359 +Check continuation, hosts, and channels. + +17 +00:01:10.000 --> 00:01:12.163 +Your runtime brings the models and tools. + +18 +00:01:14.000 --> 00:01:18.033 +Loop X. Keep the research, evidence, and +next decision together. diff --git a/docs/assets/personal-workspace/loopx-research-a.mp4 b/docs/assets/personal-workspace/loopx-research-a.mp4 new file mode 100644 index 0000000000..50f765d3a5 Binary files /dev/null and b/docs/assets/personal-workspace/loopx-research-a.mp4 differ diff --git a/docs/assets/personal-workspace/loopx-research-a.transcript.md b/docs/assets/personal-workspace/loopx-research-a.transcript.md new file mode 100644 index 0000000000..d9108d92ee --- /dev/null +++ b/docs/assets/personal-workspace/loopx-research-a.transcript.md @@ -0,0 +1,83 @@ +# LoopX 2.0 · Product flow +The workbench for long-horizon research. For enterprise teams, research organizations, and independent builders. +2.0 development preview. This is a candidate film, not a release declaration. + +[Video / 视频](https://github.com/user-attachments/assets/3b375d83-f275-41a1-bca1-94fc877efca4) · [EN VTT](loopx-research-a.en.vtt) / [SRT](loopx-research-a.en.srt) · [中文字幕](loopx-research-a.zh-CN.vtt) / [SRT](loopx-research-a.zh-CN.srt) +[English](loopx-research-a.transcript.md) · [简体中文](loopx-research-a.transcript.zh-CN.md) + + +## 00:00:00.000–00:00:07.000 · The workbench for long-horizon research. +2.0 DEVELOPMENT PREVIEW + +Loop X two point zero. A workbench for long-horizon research. + +## 00:00:07.000–00:00:18.000 · Delegate the goal. Keep its context. +ILLUSTRATION + SEPARATE RECORDED ENGINEERING RETURN + +Carry the goal, constraints, and return route. +Let the existing owner adopt the work. +Keep corrections on the task. Return the evidence. + +## 00:00:18.000–00:00:22.500 · Make research work visible. +AUTHORED WORKSPACE SCENARIO + +See work, waiting decisions, and schedules. + +## 00:00:22.500–00:00:26.500 · Make the next action explicit. +AUTHORED WORKSPACE SCENARIO + +Separate the next task from a pending approval. + +## 00:00:26.500–00:00:30.000 · Keep responsibility visible. +RECORDED WORKSPACE · ORIGINAL STATUSES RETAINED + +Keep owners and waiting work visible. + +## 00:00:30.000–00:00:35.500 · Trace the research across sessions. +RECORDED WORKSPACE · SEPARATE RESEARCH HISTORY + +Keep questions, findings, and changes of direction. + +## 00:00:35.500–00:00:45.000 · Inspect the process. Check the source. +PACKAGED UI FIXTURE · SYNTHETIC DATA + ILLUSTRATED SOURCE PATH + +Read six process judgments. Keep unknowns visible. +Trace claims to their source. Measure results and costs separately. + +## 00:00:45.000–00:00:53.400 · Read the result in its scope. +ONE HISTORICAL SOLVER RECORD · TWO CPU THREADS + +One generated-graph probe improved from forty one point two to eighty five percent. +That result applies to the duplication condition shown. + +## 00:00:53.400–00:01:00.600 · Count the cost. Preserve the counterexample. +SAME HISTORICAL RECORD · NO INDEPENDENT RERUN + +The independent control kept its accuracy. Time rose thirty two percent. +Native score gain remains unproven. + +## 00:01:00.600–00:01:06.100 · Carry the evidence into the next step. +ILLUSTRATED WORKFLOW · NO NEW TEST EXECUTION + +Keep failed paths and their reasons. Choose the next test. + +## 00:01:06.100–00:01:09.700 · Control when research continues. +AUTHORED SCENARIO · UNCONFIGURED AND STOPPED STATES + +Check continuation, hosts, and channels. + +## 00:01:09.700–00:01:13.700 · Bring your models and tools. +MECHANISM + REUSED WATERMARKED SEEDANCE 2.5 CONCEPT + +Your runtime brings the models and tools. + +## 00:01:13.700–00:01:19.200 · Research that can continue. +OPEN SOURCE · LOCAL FIRST · 2.0 DEVELOPMENT LINE + +Loop X. Keep the research, evidence, and next decision together. + +## Sources and limits +Historical record: generated full-size joint-neighborhood graphs, 30% partial-duplication probe, two CPU threads; accuracy 41.2% → 85.0%. Independent-edge accuracy unchanged, time +32%. Native score gain unproven. No independent rerun or LoopX-versus-baseline comparison. +The six process judgments are goal alignment, benefit testing, discrimination, information gain, evidence adoption and avoidable repetition. They use synthetic packaged-UI data; uncalibrated, with separate eligibility and no total score. See the [assessment contract](../../reference/trajectory-assessment.md). Citation paths and teaching choices are illustrated. Recorded workspace and engineering return are separate source cases. +English narration uses a generic local synthetic voice. Chinese subtitles translate each spoken cue. A labelled watermarked Seedance 2.5 concept insert is reused; no new remote generation, analysis or benchmark job was launched. The proposed next native test has not been run for this film. +Historical [source record](https://github.com/loopx-project/loopx/blob/c9c3adf63f9771ef95f1fcca12b11ed7de1f69eb/apps/presentation/site/public/lab/time-machine/data/graph.json). diff --git a/docs/assets/personal-workspace/loopx-research-a.transcript.zh-CN.md b/docs/assets/personal-workspace/loopx-research-a.transcript.zh-CN.md new file mode 100644 index 0000000000..3cc5b368dc --- /dev/null +++ b/docs/assets/personal-workspace/loopx-research-a.transcript.zh-CN.md @@ -0,0 +1,83 @@ +# LoopX 2.0 · 产品流程版 +长程研究工作台，面向企业、研究机构与超级个体。 +2.0 开发线候选影片，不表示已正式发布。 + +[Video / 视频](https://github.com/user-attachments/assets/3b375d83-f275-41a1-bca1-94fc877efca4) · [EN VTT](loopx-research-a.en.vtt) / [SRT](loopx-research-a.en.srt) · [中文字幕](loopx-research-a.zh-CN.vtt) / [SRT](loopx-research-a.zh-CN.srt) +[English](loopx-research-a.transcript.md) · [简体中文](loopx-research-a.transcript.zh-CN.md) + + +## 00:00:00.000–00:00:07.000 · 长程研究工作台 +2.0 DEVELOPMENT PREVIEW + +LoopX 2.0，长程研究工作台。 + +## 00:00:07.000–00:00:18.000 · 委派目标，保留上下文 +ILLUSTRATION + SEPARATE RECORDED ENGINEERING RETURN + +目标、约束和回传路径，随委派一起传递。 +由已有负责人判断并接手。 +纠正留在同一任务，证据回到原对话。 + +## 00:00:18.000–00:00:22.500 · 看清研究工作与待决事项 +AUTHORED WORKSPACE SCENARIO + +看清工作、待决策事项和日程。 + +## 00:00:22.500–00:00:26.500 · 明确下一步工作 +AUTHORED WORKSPACE SCENARIO + +把下一项任务与等待批准的事项分开。 + +## 00:00:26.500–00:00:30.000 · 保留清晰的责任归属 +RECORDED WORKSPACE · ORIGINAL STATUSES RETAINED + +看清负责人，也保留等待中的工作。 + +## 00:00:30.000–00:00:35.500 · 跨会话追溯研究过程 +RECORDED WORKSPACE · SEPARATE RESEARCH HISTORY + +保留问题、发现和研究转向。 + +## 00:00:35.500–00:00:45.000 · 分析过程，追溯证据 +PACKAGED UI FIXTURE · SYNTHETIC DATA + ILLUSTRATED SOURCE PATH + +分别看六类过程判断，保留未知项。 +沿主张追溯来源，实测结果与成本单独呈现。 + +## 00:00:45.000–00:00:53.400 · 把结果放回适用条件 +ONE HISTORICAL SOLVER RECORD · TWO CPU THREADS + +一条生成图探测记录中，准确率从41.2%升至85%。 +这个结果只适用于画面列出的重复条件。 + +## 00:00:53.400–00:01:00.600 · 计算成本，保留反例 +SAME HISTORICAL RECORD · NO INDEPENDENT RERUN + +独立边对照的准确率未变，耗时增加32%。 +原生分数收益仍未证实。 + +## 00:01:00.600–00:01:06.100 · 让证据进入下一步决策 +ILLUSTRATED WORKFLOW · NO NEW TEST EXECUTION + +保留失败路线和原因，再选择下一项检验。 + +## 00:01:06.100–00:01:09.700 · 控制研究何时继续 +AUTHORED SCENARIO · UNCONFIGURED AND STOPPED STATES + +检查续行设置、执行端和渠道。 + +## 00:01:09.700–00:01:13.700 · 使用你自己的模型和工具 +MECHANISM + REUSED WATERMARKED SEEDANCE 2.5 CONCEPT + +运行时提供模型和工具。 + +## 00:01:13.700–00:01:19.200 · 让研究有依据地继续 +OPEN SOURCE · LOCAL FIRST · 2.0 DEVELOPMENT LINE + +LoopX，让研究、证据和下一次决策衔接起来。 + +## 来源与范围 +历史记录使用完整尺寸的生成联合邻域图，在30%局部重复条件、两个 CPU 线程下，准确率从41.2%升至85.0%。独立边对照准确率不变，耗时增加32%。原生分数收益仍未证实；这不是独立复跑，也不是 LoopX 与基线的因果对照。 +六类过程判断分别是目标对齐、收益检验、区分能力、信息增益、证据采用和可避免的重复。画面使用打包 UI 的合成示例数据；判断未经科学校准，各自保留适用分母，不产生总分。定义见[轨迹分析契约](../../reference/trajectory-assessment.md)。引用路径和教学选择为机制示意。工作区历史与软件工程回传来自不同的实录案例。 +英文旁白使用本机通用合成声，中文字幕逐句对应。复用的 Seedance 2.5 概念镜头保留水印；本次没有新增远程生成、分析或 benchmark 任务。片中提出的下一步原生检验并未执行。 +[历史固定来源](https://github.com/loopx-project/loopx/blob/c9c3adf63f9771ef95f1fcca12b11ed7de1f69eb/apps/presentation/site/public/lab/time-machine/data/graph.json)。 diff --git a/docs/assets/personal-workspace/loopx-research-a.zh-CN.srt b/docs/assets/personal-workspace/loopx-research-a.zh-CN.srt new file mode 100644 index 0000000000..6b9e5f1a76 --- /dev/null +++ b/docs/assets/personal-workspace/loopx-research-a.zh-CN.srt @@ -0,0 +1,71 @@ +1 +00:00:00,300 --> 00:00:04,089 +LoopX 2.0，长程研究工作台。 + +2 +00:00:07,300 --> 00:00:10,189 +目标、约束和回传路径，随委派一起传递。 + +3 +00:00:10,369 --> 00:00:12,440 +由已有负责人判断并接手。 + +4 +00:00:12,620 --> 00:00:15,602 +纠正留在同一任务，证据回到原对话。 + +5 +00:00:18,300 --> 00:00:21,287 +看清工作、待决策事项和日程。 + +6 +00:00:22,800 --> 00:00:25,171 +把下一项任务与等待批准的事项分开。 + +7 +00:00:26,800 --> 00:00:28,667 +看清负责人，也保留等待中的工作。 + +8 +00:00:30,300 --> 00:00:33,723 +保留问题、发现和研究转向。 + +9 +00:00:35,800 --> 00:00:39,046 +分别看六类过程判断，保留未知项。 + +10 +00:00:39,226 --> 00:00:42,872 +沿主张追溯来源，实测结果与成本单独呈现。 + +11 +00:00:45,300 --> 00:00:49,877 +一条生成图探测记录中，准确率从41.2%升至85%。 + +12 +00:00:50,057 --> 00:00:53,062 +这个结果只适用于画面列出的重复条件。 + +13 +00:00:53,700 --> 00:00:58,099 +独立边对照的准确率未变，耗时增加32%。 + +14 +00:00:58,279 --> 00:01:00,236 +原生分数收益仍未证实。 + +15 +00:01:00,900 --> 00:01:04,091 +保留失败路线和原因，再选择下一项检验。 + +16 +00:01:06,400 --> 00:01:09,359 +检查续行设置、执行端和渠道。 + +17 +00:01:10,000 --> 00:01:12,163 +运行时提供模型和工具。 + +18 +00:01:14,000 --> 00:01:18,033 +LoopX，让研究、证据和下一次决策衔接起来。 diff --git a/docs/assets/personal-workspace/loopx-research-a.zh-CN.vtt b/docs/assets/personal-workspace/loopx-research-a.zh-CN.vtt new file mode 100644 index 0000000000..0a878ee37e --- /dev/null +++ b/docs/assets/personal-workspace/loopx-research-a.zh-CN.vtt @@ -0,0 +1,73 @@ +WEBVTT + +1 +00:00:00.300 --> 00:00:04.089 +LoopX 2.0，长程研究工作台。 + +2 +00:00:07.300 --> 00:00:10.189 +目标、约束和回传路径，随委派一起传递。 + +3 +00:00:10.369 --> 00:00:12.440 +由已有负责人判断并接手。 + +4 +00:00:12.620 --> 00:00:15.602 +纠正留在同一任务，证据回到原对话。 + +5 +00:00:18.300 --> 00:00:21.287 +看清工作、待决策事项和日程。 + +6 +00:00:22.800 --> 00:00:25.171 +把下一项任务与等待批准的事项分开。 + +7 +00:00:26.800 --> 00:00:28.667 +看清负责人，也保留等待中的工作。 + +8 +00:00:30.300 --> 00:00:33.723 +保留问题、发现和研究转向。 + +9 +00:00:35.800 --> 00:00:39.046 +分别看六类过程判断，保留未知项。 + +10 +00:00:39.226 --> 00:00:42.872 +沿主张追溯来源，实测结果与成本单独呈现。 + +11 +00:00:45.300 --> 00:00:49.877 +一条生成图探测记录中，准确率从41.2%升至85%。 + +12 +00:00:50.057 --> 00:00:53.062 +这个结果只适用于画面列出的重复条件。 + +13 +00:00:53.700 --> 00:00:58.099 +独立边对照的准确率未变，耗时增加32%。 + +14 +00:00:58.279 --> 00:01:00.236 +原生分数收益仍未证实。 + +15 +00:01:00.900 --> 00:01:04.091 +保留失败路线和原因，再选择下一项检验。 + +16 +00:01:06.400 --> 00:01:09.359 +检查续行设置、执行端和渠道。 + +17 +00:01:10.000 --> 00:01:12.163 +运行时提供模型和工具。 + +18 +00:01:14.000 --> 00:01:18.033 +LoopX，让研究、证据和下一次决策衔接起来。 diff --git a/docs/assets/personal-workspace/loopx-research-b.en.srt b/docs/assets/personal-workspace/loopx-research-b.en.srt new file mode 100644 index 0000000000..ea984fc2e5 --- /dev/null +++ b/docs/assets/personal-workspace/loopx-research-b.en.srt @@ -0,0 +1,89 @@ +1 +00:00:00,300 --> 00:00:04,089 +Loop X two point zero. A workbench for +long-horizon research. + +2 +00:00:07,300 --> 00:00:10,189 +Carry the goal, constraints, and return +route. + +3 +00:00:10,369 --> 00:00:12,440 +Let the existing owner adopt the work. + +4 +00:00:12,620 --> 00:00:15,602 +Keep corrections on the task. Return the +evidence. + +5 +00:00:18,300 --> 00:00:21,287 +See work, waiting decisions, and schedules. + +6 +00:00:22,800 --> 00:00:25,586 +Can a linear gate miss a useful nonlinear +model? + +7 +00:00:25,766 --> 00:00:28,330 +Keep both explanations open until the test. + +8 +00:00:29,300 --> 00:00:35,239 +On the thirty percent duplication probe, +accuracy rose from forty one point two to +eighty five percent. + +9 +00:00:37,300 --> 00:00:40,545 +On the independent control, accuracy stayed +the same. + +10 +00:00:40,725 --> 00:00:42,595 +Time increased thirty two percent. + +11 +00:00:44,800 --> 00:00:47,757 +A linear gate cannot rule out all nonlinear +headroom. + +12 +00:00:47,937 --> 00:00:51,952 +Keep the candidate eligible. Its native +score benefit is still unknown. + +13 +00:00:52,600 --> 00:00:56,348 +Keep the finding, the control, and the +reason for the next test. + +14 +00:00:58,100 --> 00:01:01,346 +Read six process judgments. Keep unknowns +visible. + +15 +00:01:01,526 --> 00:01:05,172 +Trace claims to their source. Measure +results and costs separately. + +16 +00:01:07,600 --> 00:01:10,704 +Use a matched native test. Record the full +cost. + +17 +00:01:10,884 --> 00:01:12,780 +Do not promote an untested claim. + +18 +00:01:13,600 --> 00:01:15,630 +Keep the next step with its evidence. + +19 +00:01:17,600 --> 00:01:21,633 +Loop X. Keep the research, evidence, and +next decision together. diff --git a/docs/assets/personal-workspace/loopx-research-b.en.vtt b/docs/assets/personal-workspace/loopx-research-b.en.vtt new file mode 100644 index 0000000000..5a8ea8eafb --- /dev/null +++ b/docs/assets/personal-workspace/loopx-research-b.en.vtt @@ -0,0 +1,91 @@ +WEBVTT + +1 +00:00:00.300 --> 00:00:04.089 +Loop X two point zero. A workbench for +long-horizon research. + +2 +00:00:07.300 --> 00:00:10.189 +Carry the goal, constraints, and return +route. + +3 +00:00:10.369 --> 00:00:12.440 +Let the existing owner adopt the work. + +4 +00:00:12.620 --> 00:00:15.602 +Keep corrections on the task. Return the +evidence. + +5 +00:00:18.300 --> 00:00:21.287 +See work, waiting decisions, and schedules. + +6 +00:00:22.800 --> 00:00:25.586 +Can a linear gate miss a useful nonlinear +model? + +7 +00:00:25.766 --> 00:00:28.330 +Keep both explanations open until the test. + +8 +00:00:29.300 --> 00:00:35.239 +On the thirty percent duplication probe, +accuracy rose from forty one point two to +eighty five percent. + +9 +00:00:37.300 --> 00:00:40.545 +On the independent control, accuracy stayed +the same. + +10 +00:00:40.725 --> 00:00:42.595 +Time increased thirty two percent. + +11 +00:00:44.800 --> 00:00:47.757 +A linear gate cannot rule out all nonlinear +headroom. + +12 +00:00:47.937 --> 00:00:51.952 +Keep the candidate eligible. Its native +score benefit is still unknown. + +13 +00:00:52.600 --> 00:00:56.348 +Keep the finding, the control, and the +reason for the next test. + +14 +00:00:58.100 --> 00:01:01.346 +Read six process judgments. Keep unknowns +visible. + +15 +00:01:01.526 --> 00:01:05.172 +Trace claims to their source. Measure +results and costs separately. + +16 +00:01:07.600 --> 00:01:10.704 +Use a matched native test. Record the full +cost. + +17 +00:01:10.884 --> 00:01:12.780 +Do not promote an untested claim. + +18 +00:01:13.600 --> 00:01:15.630 +Keep the next step with its evidence. + +19 +00:01:17.600 --> 00:01:21.633 +Loop X. Keep the research, evidence, and +next decision together. diff --git a/docs/assets/personal-workspace/loopx-research-b.mp4 b/docs/assets/personal-workspace/loopx-research-b.mp4 new file mode 100644 index 0000000000..c68760da93 Binary files /dev/null and b/docs/assets/personal-workspace/loopx-research-b.mp4 differ diff --git a/docs/assets/personal-workspace/loopx-research-b.transcript.md b/docs/assets/personal-workspace/loopx-research-b.transcript.md new file mode 100644 index 0000000000..aa5e72592e --- /dev/null +++ b/docs/assets/personal-workspace/loopx-research-b.transcript.md @@ -0,0 +1,80 @@ +# LoopX 2.0 · Research story +The workbench for long-horizon research. For enterprise teams, research organizations, and independent builders. +2.0 development preview. This is a candidate film, not a release declaration. + +[Video / 视频](https://github.com/user-attachments/assets/7e012d01-580e-40d0-9d94-c338f70c57d3) · [EN VTT](loopx-research-b.en.vtt) / [SRT](loopx-research-b.en.srt) · [中文字幕](loopx-research-b.zh-CN.vtt) / [SRT](loopx-research-b.zh-CN.srt) +[English](loopx-research-b.transcript.md) · [简体中文](loopx-research-b.transcript.zh-CN.md) + + +## 00:00:00.000–00:00:07.000 · The workbench for long-horizon research. +2.0 DEVELOPMENT PREVIEW + +Loop X two point zero. A workbench for long-horizon research. + +## 00:00:07.000–00:00:18.000 · Delegate the goal. Keep its context. +ILLUSTRATION + SEPARATE RECORDED ENGINEERING RETURN + +Carry the goal, constraints, and return route. +Let the existing owner adopt the work. +Keep corrections on the task. Return the evidence. + +## 00:00:18.000–00:00:22.500 · Make research work visible. +AUTHORED WORKSPACE SCENARIO + +See work, waiting decisions, and schedules. + +## 00:00:22.500–00:00:29.000 · Can a linear gate miss nonlinear headroom? +QUESTION RECONSTRUCTED FROM ONE HISTORICAL RECORD + +Can a linear gate miss a useful nonlinear model? +Keep both explanations open until the test. + +## 00:00:29.000–00:00:37.000 · The probe shows quality headroom. +GENERATED GRAPHS · 30% DUPLICATION · TWO CPU THREADS + +On the thirty percent duplication probe, accuracy rose from forty one point two to eighty five percent. + +## 00:00:37.000–00:00:44.500 · The control limits that conclusion. +SAME RECORD · INDEPENDENT-EDGE CONTROL + +On the independent control, accuracy stayed the same. +Time increased thirty two percent. + +## 00:00:44.500–00:00:52.300 · A local gain is not a native score gain. +RECORDED INTERPRETATION · NATIVE UTILITY UNKNOWN + +A linear gate cannot rule out all nonlinear headroom. +Keep the candidate eligible. Its native score benefit is still unknown. + +## 00:00:52.300–00:00:57.800 · Keep the positive and negative evidence. +SEPARATE RECORDED WORKSPACE · WORKFLOW ILLUSTRATION + +Keep the finding, the control, and the reason for the next test. + +## 00:00:57.800–00:01:07.300 · Inspect the process. Check the source. +PACKAGED UI FIXTURE · SYNTHETIC DATA + ILLUSTRATED SOURCE PATH + +Read six process judgments. Keep unknowns visible. +Trace claims to their source. Measure results and costs separately. + +## 00:01:07.300–00:01:13.300 · Propose a matched native comparison. +PROPOSED NEXT STEP · NO JOB LAUNCHED + +Use a matched native test. Record the full cost. +Do not promote an untested claim. + +## 00:01:13.300–00:01:17.300 · Keep the research ready to resume. +REUSED CONCEPT INSERT · NOT LIVE EXECUTION + +Keep the next step with its evidence. + +## 00:01:17.300–00:01:22.800 · Research that can continue. +OPEN SOURCE · LOCAL FIRST · 2.0 DEVELOPMENT LINE + +Loop X. Keep the research, evidence, and next decision together. + +## Sources and limits +Historical record: generated full-size joint-neighborhood graphs, 30% partial-duplication probe, two CPU threads; accuracy 41.2% → 85.0%. Independent-edge accuracy unchanged, time +32%. Native score gain unproven. No independent rerun or LoopX-versus-baseline comparison. +The six process judgments are goal alignment, benefit testing, discrimination, information gain, evidence adoption and avoidable repetition. They use synthetic packaged-UI data; uncalibrated, with separate eligibility and no total score. See the [assessment contract](../../reference/trajectory-assessment.md). Citation paths and teaching choices are illustrated. Recorded workspace and engineering return are separate source cases. +English narration uses a generic local synthetic voice. Chinese subtitles translate each spoken cue. A labelled watermarked Seedance 2.5 concept insert is reused; no new remote generation, analysis or benchmark job was launched. The proposed next native test has not been run for this film. +Historical [source record](https://github.com/loopx-project/loopx/blob/c9c3adf63f9771ef95f1fcca12b11ed7de1f69eb/apps/presentation/site/public/lab/time-machine/data/graph.json). diff --git a/docs/assets/personal-workspace/loopx-research-b.transcript.zh-CN.md b/docs/assets/personal-workspace/loopx-research-b.transcript.zh-CN.md new file mode 100644 index 0000000000..ad0d8ed7f2 --- /dev/null +++ b/docs/assets/personal-workspace/loopx-research-b.transcript.zh-CN.md @@ -0,0 +1,80 @@ +# LoopX 2.0 · 研究故事版 +长程研究工作台，面向企业、研究机构与超级个体。 +2.0 开发线候选影片，不表示已正式发布。 + +[Video / 视频](https://github.com/user-attachments/assets/7e012d01-580e-40d0-9d94-c338f70c57d3) · [EN VTT](loopx-research-b.en.vtt) / [SRT](loopx-research-b.en.srt) · [中文字幕](loopx-research-b.zh-CN.vtt) / [SRT](loopx-research-b.zh-CN.srt) +[English](loopx-research-b.transcript.md) · [简体中文](loopx-research-b.transcript.zh-CN.md) + + +## 00:00:00.000–00:00:07.000 · 长程研究工作台 +2.0 DEVELOPMENT PREVIEW + +LoopX 2.0，长程研究工作台。 + +## 00:00:07.000–00:00:18.000 · 委派目标，保留上下文 +ILLUSTRATION + SEPARATE RECORDED ENGINEERING RETURN + +目标、约束和回传路径，随委派一起传递。 +由已有负责人判断并接手。 +纠正留在同一任务，证据回到原对话。 + +## 00:00:18.000–00:00:22.500 · 看清研究工作与待决事项 +AUTHORED WORKSPACE SCENARIO + +看清工作、待决策事项和日程。 + +## 00:00:22.500–00:00:29.000 · 线性门槛会不会漏掉非线性能力？ +QUESTION RECONSTRUCTED FROM ONE HISTORICAL RECORD + +线性门槛会不会漏掉有用的非线性模型？ +在检验前，保留两种解释。 + +## 00:00:29.000–00:00:37.000 · 探测显示了质量提升空间 +GENERATED GRAPHS · 30% DUPLICATION · TWO CPU THREADS + +在30%重复条件的探测中，准确率从41.2%升至85%。 + +## 00:00:37.000–00:00:44.500 · 对照限定了结论的范围 +SAME RECORD · INDEPENDENT-EDGE CONTROL + +在独立边对照上，准确率没有变化。 +耗时增加了32%。 + +## 00:00:44.500–00:00:52.300 · 局部提升不等于原生分数收益 +RECORDED INTERPRETATION · NATIVE UTILITY UNKNOWN + +线性门槛不能排除全部非线性提升空间。 +保留候选资格，原生分数收益仍然未知。 + +## 00:00:52.300–00:00:57.800 · 把正向发现与反例一起保留 +SEPARATE RECORDED WORKSPACE · WORKFLOW ILLUSTRATION + +同时保留发现、对照和下一项检验的理由。 + +## 00:00:57.800–00:01:07.300 · 分析过程，追溯证据 +PACKAGED UI FIXTURE · SYNTHETIC DATA + ILLUSTRATED SOURCE PATH + +分别看六类过程判断，保留未知项。 +沿主张追溯来源，实测结果与成本单独呈现。 + +## 00:01:07.300–00:01:13.300 · 提出条件匹配的原生比较 +PROPOSED NEXT STEP · NO JOB LAUNCHED + +用条件匹配的原生检验，并记录完整成本。 +未经检验的主张，不升级为收益结论。 + +## 00:01:13.300–00:01:17.300 · 让研究可以继续接续 +REUSED CONCEPT INSERT · NOT LIVE EXECUTION + +把下一步与它的证据放在一起。 + +## 00:01:17.300–00:01:22.800 · 让研究有依据地继续 +OPEN SOURCE · LOCAL FIRST · 2.0 DEVELOPMENT LINE + +LoopX，让研究、证据和下一次决策衔接起来。 + +## 来源与范围 +历史记录使用完整尺寸的生成联合邻域图，在30%局部重复条件、两个 CPU 线程下，准确率从41.2%升至85.0%。独立边对照准确率不变，耗时增加32%。原生分数收益仍未证实；这不是独立复跑，也不是 LoopX 与基线的因果对照。 +六类过程判断分别是目标对齐、收益检验、区分能力、信息增益、证据采用和可避免的重复。画面使用打包 UI 的合成示例数据；判断未经科学校准，各自保留适用分母，不产生总分。定义见[轨迹分析契约](../../reference/trajectory-assessment.md)。引用路径和教学选择为机制示意。工作区历史与软件工程回传来自不同的实录案例。 +英文旁白使用本机通用合成声，中文字幕逐句对应。复用的 Seedance 2.5 概念镜头保留水印；本次没有新增远程生成、分析或 benchmark 任务。片中提出的下一步原生检验并未执行。 +[历史固定来源](https://github.com/loopx-project/loopx/blob/c9c3adf63f9771ef95f1fcca12b11ed7de1f69eb/apps/presentation/site/public/lab/time-machine/data/graph.json)。 diff --git a/docs/assets/personal-workspace/loopx-research-b.zh-CN.srt b/docs/assets/personal-workspace/loopx-research-b.zh-CN.srt new file mode 100644 index 0000000000..17d8fbd7f3 --- /dev/null +++ b/docs/assets/personal-workspace/loopx-research-b.zh-CN.srt @@ -0,0 +1,75 @@ +1 +00:00:00,300 --> 00:00:04,089 +LoopX 2.0，长程研究工作台。 + +2 +00:00:07,300 --> 00:00:10,189 +目标、约束和回传路径，随委派一起传递。 + +3 +00:00:10,369 --> 00:00:12,440 +由已有负责人判断并接手。 + +4 +00:00:12,620 --> 00:00:15,602 +纠正留在同一任务，证据回到原对话。 + +5 +00:00:18,300 --> 00:00:21,287 +看清工作、待决策事项和日程。 + +6 +00:00:22,800 --> 00:00:25,586 +线性门槛会不会漏掉有用的非线性模型？ + +7 +00:00:25,766 --> 00:00:28,330 +在检验前，保留两种解释。 + +8 +00:00:29,300 --> 00:00:35,239 +在30%重复条件的探测中，准确率从41.2%升至85%。 + +9 +00:00:37,300 --> 00:00:40,545 +在独立边对照上，准确率没有变化。 + +10 +00:00:40,725 --> 00:00:42,595 +耗时增加了32%。 + +11 +00:00:44,800 --> 00:00:47,757 +线性门槛不能排除全部非线性提升空间。 + +12 +00:00:47,937 --> 00:00:51,952 +保留候选资格，原生分数收益仍然未知。 + +13 +00:00:52,600 --> 00:00:56,348 +同时保留发现、对照和下一项检验的理由。 + +14 +00:00:58,100 --> 00:01:01,346 +分别看六类过程判断，保留未知项。 + +15 +00:01:01,526 --> 00:01:05,172 +沿主张追溯来源，实测结果与成本单独呈现。 + +16 +00:01:07,600 --> 00:01:10,704 +用条件匹配的原生检验，并记录完整成本。 + +17 +00:01:10,884 --> 00:01:12,780 +未经检验的主张，不升级为收益结论。 + +18 +00:01:13,600 --> 00:01:15,630 +把下一步与它的证据放在一起。 + +19 +00:01:17,600 --> 00:01:21,633 +LoopX，让研究、证据和下一次决策衔接起来。 diff --git a/docs/assets/personal-workspace/loopx-research-b.zh-CN.vtt b/docs/assets/personal-workspace/loopx-research-b.zh-CN.vtt new file mode 100644 index 0000000000..edbc11da0d --- /dev/null +++ b/docs/assets/personal-workspace/loopx-research-b.zh-CN.vtt @@ -0,0 +1,77 @@ +WEBVTT + +1 +00:00:00.300 --> 00:00:04.089 +LoopX 2.0，长程研究工作台。 + +2 +00:00:07.300 --> 00:00:10.189 +目标、约束和回传路径，随委派一起传递。 + +3 +00:00:10.369 --> 00:00:12.440 +由已有负责人判断并接手。 + +4 +00:00:12.620 --> 00:00:15.602 +纠正留在同一任务，证据回到原对话。 + +5 +00:00:18.300 --> 00:00:21.287 +看清工作、待决策事项和日程。 + +6 +00:00:22.800 --> 00:00:25.586 +线性门槛会不会漏掉有用的非线性模型？ + +7 +00:00:25.766 --> 00:00:28.330 +在检验前，保留两种解释。 + +8 +00:00:29.300 --> 00:00:35.239 +在30%重复条件的探测中，准确率从41.2%升至85%。 + +9 +00:00:37.300 --> 00:00:40.545 +在独立边对照上，准确率没有变化。 + +10 +00:00:40.725 --> 00:00:42.595 +耗时增加了32%。 + +11 +00:00:44.800 --> 00:00:47.757 +线性门槛不能排除全部非线性提升空间。 + +12 +00:00:47.937 --> 00:00:51.952 +保留候选资格，原生分数收益仍然未知。 + +13 +00:00:52.600 --> 00:00:56.348 +同时保留发现、对照和下一项检验的理由。 + +14 +00:00:58.100 --> 00:01:01.346 +分别看六类过程判断，保留未知项。 + +15 +00:01:01.526 --> 00:01:05.172 +沿主张追溯来源，实测结果与成本单独呈现。 + +16 +00:01:07.600 --> 00:01:10.704 +用条件匹配的原生检验，并记录完整成本。 + +17 +00:01:10.884 --> 00:01:12.780 +未经检验的主张，不升级为收益结论。 + +18 +00:01:13.600 --> 00:01:15.630 +把下一步与它的证据放在一起。 + +19 +00:01:17.600 --> 00:01:21.633 +LoopX，让研究、证据和下一次决策衔接起来。 diff --git a/docs/assets/personal-workspace/loopx-research-c.en.srt b/docs/assets/personal-workspace/loopx-research-c.en.srt new file mode 100644 index 0000000000..95d5e7254e --- /dev/null +++ b/docs/assets/personal-workspace/loopx-research-c.en.srt @@ -0,0 +1,86 @@ +1 +00:00:00,300 --> 00:00:04,089 +Loop X two point zero. A workbench for +long-horizon research. + +2 +00:00:07,300 --> 00:00:10,189 +Carry the goal, constraints, and return +route. + +3 +00:00:10,369 --> 00:00:12,440 +Let the existing owner adopt the work. + +4 +00:00:12,620 --> 00:00:15,602 +Keep corrections on the task. Return the +evidence. + +5 +00:00:18,300 --> 00:00:21,287 +See work, waiting decisions, and schedules. + +6 +00:00:22,800 --> 00:00:25,374 +A claim applies only to its tested +condition. + +7 +00:00:25,554 --> 00:00:28,443 +Check the control. Keep unknown benefits +unknown. + +8 +00:00:29,800 --> 00:00:33,049 +This probe improved. The native score is not +established. + +9 +00:00:35,800 --> 00:00:39,295 +Pause here. Reread the headline, or inspect +the control? + +10 +00:00:42,800 --> 00:00:45,570 +Repeating the headline adds no recorded +evidence here. + +11 +00:00:45,750 --> 00:00:50,174 +The control keeps accuracy unchanged, with +thirty two percent more time. + +12 +00:00:51,800 --> 00:00:55,659 +Both paths use the same record. Native +benefit remains unknown. + +13 +00:00:57,300 --> 00:01:00,677 +Keep the question, the evidence, and your +next decision. + +14 +00:01:02,300 --> 00:01:05,546 +Read six process judgments. Keep unknowns +visible. + +15 +00:01:05,726 --> 00:01:09,372 +Trace claims to their source. Measure +results and costs separately. + +16 +00:01:11,800 --> 00:01:14,933 +Choose a matched test. Record its result and +cost. + +17 +00:01:17,800 --> 00:01:19,830 +Keep the next step with its evidence. + +18 +00:01:21,800 --> 00:01:25,833 +Loop X. Keep the research, evidence, and +next decision together. diff --git a/docs/assets/personal-workspace/loopx-research-c.en.vtt b/docs/assets/personal-workspace/loopx-research-c.en.vtt new file mode 100644 index 0000000000..10b03f2b49 --- /dev/null +++ b/docs/assets/personal-workspace/loopx-research-c.en.vtt @@ -0,0 +1,88 @@ +WEBVTT + +1 +00:00:00.300 --> 00:00:04.089 +Loop X two point zero. A workbench for +long-horizon research. + +2 +00:00:07.300 --> 00:00:10.189 +Carry the goal, constraints, and return +route. + +3 +00:00:10.369 --> 00:00:12.440 +Let the existing owner adopt the work. + +4 +00:00:12.620 --> 00:00:15.602 +Keep corrections on the task. Return the +evidence. + +5 +00:00:18.300 --> 00:00:21.287 +See work, waiting decisions, and schedules. + +6 +00:00:22.800 --> 00:00:25.374 +A claim applies only to its tested +condition. + +7 +00:00:25.554 --> 00:00:28.443 +Check the control. Keep unknown benefits +unknown. + +8 +00:00:29.800 --> 00:00:33.049 +This probe improved. The native score is not +established. + +9 +00:00:35.800 --> 00:00:39.295 +Pause here. Reread the headline, or inspect +the control? + +10 +00:00:42.800 --> 00:00:45.570 +Repeating the headline adds no recorded +evidence here. + +11 +00:00:45.750 --> 00:00:50.174 +The control keeps accuracy unchanged, with +thirty two percent more time. + +12 +00:00:51.800 --> 00:00:55.659 +Both paths use the same record. Native +benefit remains unknown. + +13 +00:00:57.300 --> 00:01:00.677 +Keep the question, the evidence, and your +next decision. + +14 +00:01:02.300 --> 00:01:05.546 +Read six process judgments. Keep unknowns +visible. + +15 +00:01:05.726 --> 00:01:09.372 +Trace claims to their source. Measure +results and costs separately. + +16 +00:01:11.800 --> 00:01:14.933 +Choose a matched test. Record its result and +cost. + +17 +00:01:17.800 --> 00:01:19.830 +Keep the next step with its evidence. + +18 +00:01:21.800 --> 00:01:25.833 +Loop X. Keep the research, evidence, and +next decision together. diff --git a/docs/assets/personal-workspace/loopx-research-c.mp4 b/docs/assets/personal-workspace/loopx-research-c.mp4 new file mode 100644 index 0000000000..55970c6aa2 Binary files /dev/null and b/docs/assets/personal-workspace/loopx-research-c.mp4 differ diff --git a/docs/assets/personal-workspace/loopx-research-c.transcript.md b/docs/assets/personal-workspace/loopx-research-c.transcript.md new file mode 100644 index 0000000000..c3820bb118 --- /dev/null +++ b/docs/assets/personal-workspace/loopx-research-c.transcript.md @@ -0,0 +1,83 @@ +# LoopX 2.0 · Predict and compare +The workbench for long-horizon research. For enterprise teams, research organizations, and independent builders. +2.0 development preview. This is a candidate film, not a release declaration. + +[Video / 视频](https://github.com/user-attachments/assets/0eb29457-4244-4ee4-9a34-ce562074d46c) · [EN VTT](loopx-research-c.en.vtt) / [SRT](loopx-research-c.en.srt) · [中文字幕](loopx-research-c.zh-CN.vtt) / [SRT](loopx-research-c.zh-CN.srt) +[English](loopx-research-c.transcript.md) · [简体中文](loopx-research-c.transcript.zh-CN.md) + + +## 00:00:00.000–00:00:07.000 · The workbench for long-horizon research. +2.0 DEVELOPMENT PREVIEW + +Loop X two point zero. A workbench for long-horizon research. + +## 00:00:07.000–00:00:18.000 · Delegate the goal. Keep its context. +ILLUSTRATION + SEPARATE RECORDED ENGINEERING RETURN + +Carry the goal, constraints, and return route. +Let the existing owner adopt the work. +Keep corrections on the task. Return the evidence. + +## 00:00:18.000–00:00:22.500 · Make research work visible. +AUTHORED WORKSPACE SCENARIO + +See work, waiting decisions, and schedules. + +## 00:00:22.500–00:00:29.500 · One record. One set of rules. +TEACHING ILLUSTRATION · NO SIMULATED BENCHMARK SCORE + +A claim applies only to its tested condition. +Check the control. Keep unknown benefits unknown. + +## 00:00:29.500–00:00:35.500 · A probe reports eighty five percent. +GENERATED GRAPHS · 30% DUPLICATION · TWO CPU THREADS + +This probe improved. The native score is not established. + +## 00:00:35.500–00:00:42.500 · What would you check next? +PAUSE TO PREDICT · ILLUSTRATED CHOICES + +Pause here. Reread the headline, or inspect the control? + +## 00:00:42.500–00:00:51.500 · Different reading. The same underlying evidence. +ILLUSTRATED INFORMATION PATHS · NO NEW TEST EXECUTED + +Repeating the headline adds no recorded evidence here. +The control keeps accuracy unchanged, with thirty two percent more time. + +## 00:00:51.500–00:00:57.000 · Neither path proves a native score gain. +SHARED RULE · THE UNKNOWN RESULT IS NOT CHANGED + +Both paths use the same record. Native benefit remains unknown. + +## 00:00:57.000–00:01:02.000 · Keep the question and the decision. +SEPARATE RECORDED WORKSPACE · WORKFLOW ILLUSTRATION + +Keep the question, the evidence, and your next decision. + +## 00:01:02.000–00:01:11.500 · Inspect the process. Check the source. +PACKAGED UI FIXTURE · SYNTHETIC DATA + ILLUSTRATED SOURCE PATH + +Read six process judgments. Keep unknowns visible. +Trace claims to their source. Measure results and costs separately. + +## 00:01:11.500–00:01:17.500 · Choose a test that can change the decision. +PROPOSED MATCHED NATIVE TEST · NO JOB LAUNCHED + +Choose a matched test. Record its result and cost. + +## 00:01:17.500–00:01:21.500 · Continue with the evidence attached. +REUSED CONCEPT INSERT · NOT LIVE EXECUTION + +Keep the next step with its evidence. + +## 00:01:21.500–00:01:27.000 · Research that can continue. +OPEN SOURCE · LOCAL FIRST · 2.0 DEVELOPMENT LINE + +Loop X. Keep the research, evidence, and next decision together. + +## Sources and limits +Historical record: generated full-size joint-neighborhood graphs, 30% partial-duplication probe, two CPU threads; accuracy 41.2% → 85.0%. Independent-edge accuracy unchanged, time +32%. Native score gain unproven. No independent rerun or LoopX-versus-baseline comparison. +The six process judgments are goal alignment, benefit testing, discrimination, information gain, evidence adoption and avoidable repetition. They use synthetic packaged-UI data; uncalibrated, with separate eligibility and no total score. See the [assessment contract](../../reference/trajectory-assessment.md). Citation paths and teaching choices are illustrated. Recorded workspace and engineering return are separate source cases. +English narration uses a generic local synthetic voice. Chinese subtitles translate each spoken cue. A labelled watermarked Seedance 2.5 concept insert is reused; no new remote generation, analysis or benchmark job was launched. The proposed next native test has not been run for this film. +Historical [source record](https://github.com/loopx-project/loopx/blob/c9c3adf63f9771ef95f1fcca12b11ed7de1f69eb/apps/presentation/site/public/lab/time-machine/data/graph.json). diff --git a/docs/assets/personal-workspace/loopx-research-c.transcript.zh-CN.md b/docs/assets/personal-workspace/loopx-research-c.transcript.zh-CN.md new file mode 100644 index 0000000000..e82e2dbcb8 --- /dev/null +++ b/docs/assets/personal-workspace/loopx-research-c.transcript.zh-CN.md @@ -0,0 +1,83 @@ +# LoopX 2.0 · 机制推演版 +长程研究工作台，面向企业、研究机构与超级个体。 +2.0 开发线候选影片，不表示已正式发布。 + +[Video / 视频](https://github.com/user-attachments/assets/0eb29457-4244-4ee4-9a34-ce562074d46c) · [EN VTT](loopx-research-c.en.vtt) / [SRT](loopx-research-c.en.srt) · [中文字幕](loopx-research-c.zh-CN.vtt) / [SRT](loopx-research-c.zh-CN.srt) +[English](loopx-research-c.transcript.md) · [简体中文](loopx-research-c.transcript.zh-CN.md) + + +## 00:00:00.000–00:00:07.000 · 长程研究工作台 +2.0 DEVELOPMENT PREVIEW + +LoopX 2.0，长程研究工作台。 + +## 00:00:07.000–00:00:18.000 · 委派目标，保留上下文 +ILLUSTRATION + SEPARATE RECORDED ENGINEERING RETURN + +目标、约束和回传路径，随委派一起传递。 +由已有负责人判断并接手。 +纠正留在同一任务，证据回到原对话。 + +## 00:00:18.000–00:00:22.500 · 看清研究工作与待决事项 +AUTHORED WORKSPACE SCENARIO + +看清工作、待决策事项和日程。 + +## 00:00:22.500–00:00:29.500 · 同一份记录，同一套判断规则 +TEACHING ILLUSTRATION · NO SIMULATED BENCHMARK SCORE + +主张只适用于已检验的条件。 +检查对照，收益未知就保留未知。 + +## 00:00:29.500–00:00:35.500 · 一条探测记录报告了85%的准确率 +GENERATED GRAPHS · 30% DUPLICATION · TWO CPU THREADS + +这条探测有提升，原生分数收益尚未建立。 + +## 00:00:35.500–00:00:42.500 · 看到新证据，你会怎样继续？ +PAUSE TO PREDICT · ILLUSTRATED CHOICES + +在这里暂停：再读正向结果，还是检查对照？ + +## 00:00:42.500–00:00:51.500 · 读法不同，底层证据相同 +ILLUSTRATED INFORMATION PATHS · NO NEW TEST EXECUTED + +在这段推演中，重复标题没有增加已记录证据。 +对照的准确率未变，耗时增加32%。 + +## 00:00:51.500–00:00:57.000 · 两条路径都未证明原生分数收益 +SHARED RULE · THE UNKNOWN RESULT IS NOT CHANGED + +两条路径使用同一记录，原生收益仍然未知。 + +## 00:00:57.000–00:01:02.000 · 留下问题，也留下决策依据 +SEPARATE RECORDED WORKSPACE · WORKFLOW ILLUSTRATION + +保留问题、证据和下一次决策。 + +## 00:01:02.000–00:01:11.500 · 分析过程，追溯证据 +PACKAGED UI FIXTURE · SYNTHETIC DATA + ILLUSTRATED SOURCE PATH + +分别看六类过程判断，保留未知项。 +沿主张追溯来源，实测结果与成本单独呈现。 + +## 00:01:11.500–00:01:17.500 · 选择能改变判断的检验 +PROPOSED MATCHED NATIVE TEST · NO JOB LAUNCHED + +选择条件匹配的检验，记录结果与成本。 + +## 00:01:17.500–00:01:21.500 · 带着证据继续研究 +REUSED CONCEPT INSERT · NOT LIVE EXECUTION + +把下一步与它的证据放在一起。 + +## 00:01:21.500–00:01:27.000 · 让研究有依据地继续 +OPEN SOURCE · LOCAL FIRST · 2.0 DEVELOPMENT LINE + +LoopX，让研究、证据和下一次决策衔接起来。 + +## 来源与范围 +历史记录使用完整尺寸的生成联合邻域图，在30%局部重复条件、两个 CPU 线程下，准确率从41.2%升至85.0%。独立边对照准确率不变，耗时增加32%。原生分数收益仍未证实；这不是独立复跑，也不是 LoopX 与基线的因果对照。 +六类过程判断分别是目标对齐、收益检验、区分能力、信息增益、证据采用和可避免的重复。画面使用打包 UI 的合成示例数据；判断未经科学校准，各自保留适用分母，不产生总分。定义见[轨迹分析契约](../../reference/trajectory-assessment.md)。引用路径和教学选择为机制示意。工作区历史与软件工程回传来自不同的实录案例。 +英文旁白使用本机通用合成声，中文字幕逐句对应。复用的 Seedance 2.5 概念镜头保留水印；本次没有新增远程生成、分析或 benchmark 任务。片中提出的下一步原生检验并未执行。 +[历史固定来源](https://github.com/loopx-project/loopx/blob/c9c3adf63f9771ef95f1fcca12b11ed7de1f69eb/apps/presentation/site/public/lab/time-machine/data/graph.json)。 diff --git a/docs/assets/personal-workspace/loopx-research-c.zh-CN.srt b/docs/assets/personal-workspace/loopx-research-c.zh-CN.srt new file mode 100644 index 0000000000..6c7f00a153 --- /dev/null +++ b/docs/assets/personal-workspace/loopx-research-c.zh-CN.srt @@ -0,0 +1,71 @@ +1 +00:00:00,300 --> 00:00:04,089 +LoopX 2.0，长程研究工作台。 + +2 +00:00:07,300 --> 00:00:10,189 +目标、约束和回传路径，随委派一起传递。 + +3 +00:00:10,369 --> 00:00:12,440 +由已有负责人判断并接手。 + +4 +00:00:12,620 --> 00:00:15,602 +纠正留在同一任务，证据回到原对话。 + +5 +00:00:18,300 --> 00:00:21,287 +看清工作、待决策事项和日程。 + +6 +00:00:22,800 --> 00:00:25,374 +主张只适用于已检验的条件。 + +7 +00:00:25,554 --> 00:00:28,443 +检查对照，收益未知就保留未知。 + +8 +00:00:29,800 --> 00:00:33,049 +这条探测有提升，原生分数收益尚未建立。 + +9 +00:00:35,800 --> 00:00:39,295 +在这里暂停：再读正向结果，还是检查对照？ + +10 +00:00:42,800 --> 00:00:45,570 +在这段推演中，重复标题没有增加已记录证据。 + +11 +00:00:45,750 --> 00:00:50,174 +对照的准确率未变，耗时增加32%。 + +12 +00:00:51,800 --> 00:00:55,659 +两条路径使用同一记录，原生收益仍然未知。 + +13 +00:00:57,300 --> 00:01:00,677 +保留问题、证据和下一次决策。 + +14 +00:01:02,300 --> 00:01:05,546 +分别看六类过程判断，保留未知项。 + +15 +00:01:05,726 --> 00:01:09,372 +沿主张追溯来源，实测结果与成本单独呈现。 + +16 +00:01:11,800 --> 00:01:14,933 +选择条件匹配的检验，记录结果与成本。 + +17 +00:01:17,800 --> 00:01:19,830 +把下一步与它的证据放在一起。 + +18 +00:01:21,800 --> 00:01:25,833 +LoopX，让研究、证据和下一次决策衔接起来。 diff --git a/docs/assets/personal-workspace/loopx-research-c.zh-CN.vtt b/docs/assets/personal-workspace/loopx-research-c.zh-CN.vtt new file mode 100644 index 0000000000..7f529e7ea7 --- /dev/null +++ b/docs/assets/personal-workspace/loopx-research-c.zh-CN.vtt @@ -0,0 +1,73 @@ +WEBVTT + +1 +00:00:00.300 --> 00:00:04.089 +LoopX 2.0，长程研究工作台。 + +2 +00:00:07.300 --> 00:00:10.189 +目标、约束和回传路径，随委派一起传递。 + +3 +00:00:10.369 --> 00:00:12.440 +由已有负责人判断并接手。 + +4 +00:00:12.620 --> 00:00:15.602 +纠正留在同一任务，证据回到原对话。 + +5 +00:00:18.300 --> 00:00:21.287 +看清工作、待决策事项和日程。 + +6 +00:00:22.800 --> 00:00:25.374 +主张只适用于已检验的条件。 + +7 +00:00:25.554 --> 00:00:28.443 +检查对照，收益未知就保留未知。 + +8 +00:00:29.800 --> 00:00:33.049 +这条探测有提升，原生分数收益尚未建立。 + +9 +00:00:35.800 --> 00:00:39.295 +在这里暂停：再读正向结果，还是检查对照？ + +10 +00:00:42.800 --> 00:00:45.570 +在这段推演中，重复标题没有增加已记录证据。 + +11 +00:00:45.750 --> 00:00:50.174 +对照的准确率未变，耗时增加32%。 + +12 +00:00:51.800 --> 00:00:55.659 +两条路径使用同一记录，原生收益仍然未知。 + +13 +00:00:57.300 --> 00:01:00.677 +保留问题、证据和下一次决策。 + +14 +00:01:02.300 --> 00:01:05.546 +分别看六类过程判断，保留未知项。 + +15 +00:01:05.726 --> 00:01:09.372 +沿主张追溯来源，实测结果与成本单独呈现。 + +16 +00:01:11.800 --> 00:01:14.933 +选择条件匹配的检验，记录结果与成本。 + +17 +00:01:17.800 --> 00:01:19.830 +把下一步与它的证据放在一起。 + +18 +00:01:21.800 --> 00:01:25.833 +LoopX，让研究、证据和下一次决策衔接起来。 diff --git a/docs/assets/personal-workspace/loopx-research-films.md b/docs/assets/personal-workspace/loopx-research-films.md new file mode 100644 index 0000000000..06909306d7 --- /dev/null +++ b/docs/assets/personal-workspace/loopx-research-films.md @@ -0,0 +1,27 @@ +# LoopX 2.0 · Three research workbench films + +The workbench for long-horizon research. For enterprise teams, research organizations, and independent builders. + +2.0 development preview. These are candidates for selection, not a release declaration. + +| Version | What it teaches | Video and transcript | +| --- | --- | --- | +| A · Product flow · 79.2s | Follow a goal through delegation, research history, evidence, costs and the next action. | [MP4](https://github.com/user-attachments/assets/3b375d83-f275-41a1-bca1-94fc877efca4) · [EN](loopx-research-a.transcript.md) · [中文](loopx-research-a.transcript.zh-CN.md) | +| B · Research story · 82.8s | One graph-classification question: keep a nonlinear candidate eligible while retaining its counterexample and extra cost. | [MP4](https://github.com/user-attachments/assets/7e012d01-580e-40d0-9d94-c338f70c57d3) · [EN](loopx-research-b.transcript.md) · [中文](loopx-research-b.transcript.zh-CN.md) | +| C · Predict and compare · 87s | Pause to predict; compare two information paths over the same record. No simulated experiment. | [MP4](https://github.com/user-attachments/assets/0eb29457-4244-4ee4-9a34-ce562074d46c) · [EN](loopx-research-c.transcript.md) · [中文](loopx-research-c.transcript.zh-CN.md) | + +Each film has English narration, Chinese on-screen subtitles, EN/ZH VTT and SRT files, and a transcript. Load the matching sidecar in a player that supports external subtitles. Use fullscreen or the transcript for dense UI details. + +![三版编码关键帧 / Encoded keyframes](research-workbench-readme/film-candidates-keyframes.png) + +## Sources and interpretation + +- Delegation diagrams use a separate recorded engineering return. Research history is another recorded case; the edit does not claim they are one execution. +- Kanban and controls use a published workspace scenario. Trajectory analysis uses synthetic packaged-UI data. Its six process judgments are uncalibrated and do not form a total score. Citation paths are illustrated. +- The graph record used two CPU threads and generated full-size joint-neighborhood graphs at30% partial duplication: accuracy41.2%→85.0%. The independent-edge control kept its accuracy while time rose32%. This supports testing nonlinear candidates; native score gain remains unknown. It is not a causal LoopX-versus-baseline result. +- The watermarked Seedance2.5 concept insert reuses earlier material; new diagrams are drawn in code. No new remote generation or benchmark execution. +- Enterprise and research organizations describe the intended audience, not verified deployment, security, SLA or multitenancy. Historical1.0.3 LHTB results retain their original version. + +[Historical fixed source / 历史固定来源](https://github.com/loopx-project/loopx/blob/c9c3adf63f9771ef95f1fcca12b11ed7de1f69eb/apps/presentation/site/public/lab/time-machine/data/graph.json) · [Assessment contract / 分析契约](../../reference/trajectory-assessment.md) + +[Original58s film / 保留的原58秒影片](loopx-research-workbench.mp4) · [Original transcript / 原文字稿](loopx-research-workbench.transcript.md) diff --git a/docs/assets/personal-workspace/loopx-research-films.zh-CN.md b/docs/assets/personal-workspace/loopx-research-films.zh-CN.md new file mode 100644 index 0000000000..83e1407f9b --- /dev/null +++ b/docs/assets/personal-workspace/loopx-research-films.zh-CN.md @@ -0,0 +1,27 @@ +# LoopX 2.0 · 三版研究工作台影片 + +长程研究工作台，面向企业、研究机构与超级个体。 + +2.0 开发预览，候选待选，不表示已正式发布。 + +| 版本 | 讲什么 | 观看与文字稿 | +| --- | --- | --- | +| A · 产品流程 · 79.2s | 从目标和委派，经过研究历史、证据与成本，到下一步。 | [MP4](https://github.com/user-attachments/assets/3b375d83-f275-41a1-bca1-94fc877efca4) · [EN](loopx-research-a.transcript.md) · [中文](loopx-research-a.transcript.zh-CN.md) | +| B · 研究故事 · 82.8s | 一个图分类问题：保留非线性候选，也保留反例与额外成本。 | [MP4](https://github.com/user-attachments/assets/7e012d01-580e-40d0-9d94-c338f70c57d3) · [EN](loopx-research-b.transcript.md) · [中文](loopx-research-b.transcript.zh-CN.md) | +| C · 机制推演 · 87s | 暂停预测，再比较同一记录的两条信息路径；没有新模拟实验。 | [MP4](https://github.com/user-attachments/assets/0eb29457-4244-4ee4-9a34-ce562074d46c) · [EN](loopx-research-c.transcript.md) · [中文](loopx-research-c.transcript.zh-CN.md) | + +每版均有英文旁白、画面中文字幕、中英文 VTT/SRT 和逐句文字稿。播放器支持外部字幕时，可加载对应 VTT 或 SRT。密集界面建议全屏或配合文字稿阅读。 + +![三版编码关键帧 / Encoded keyframes](research-workbench-readme/film-candidates-keyframes.png) + +## 素材与判断边界 + +- 委派图示结合单独的软件工程实录回传；研究历史是另一份实录，不能拼成同一真实执行。 +- 看板和控制设置使用已公开的工作区场景。轨迹分析使用打包 UI 的合成数据；六类过程判断未经科学校准，不产生总分。引用路径为示意。 +- 图分类记录来自两个 CPU 线程、完整尺寸生成联合邻域图的30%局部重复条件：准确率41.2%→85.0%。独立边对照准确率未变、耗时+32%。它支持继续检验非线性候选，原生分数收益仍未知；不构成 LoopX 与基线的因果收益。 +- 带水印的 Seedance 2.5 概念镜头复用旧素材；其余新增画面由代码绘制。没有新增远程生成或 benchmark 执行。 +- 企业和机构是受众定位，不表示已验证企业部署、安全、SLA或多租户；历史1.0.3 LHTB结果不改称2.0实测。 + +[Historical fixed source / 历史固定来源](https://github.com/loopx-project/loopx/blob/c9c3adf63f9771ef95f1fcca12b11ed7de1f69eb/apps/presentation/site/public/lab/time-machine/data/graph.json) · [Assessment contract / 分析契约](../../reference/trajectory-assessment.md) + +[Original58s film / 保留的原58秒影片](loopx-research-workbench.mp4) · [Original transcript / 原文字稿](loopx-research-workbench.transcript.zh-CN.md) diff --git a/docs/assets/personal-workspace/loopx-research-workbench.en.srt b/docs/assets/personal-workspace/loopx-research-workbench.en.srt new file mode 100644 index 0000000000..860e8fbe04 --- /dev/null +++ b/docs/assets/personal-workspace/loopx-research-workbench.en.srt @@ -0,0 +1,87 @@ +1 +00:00:00,080 --> 00:00:02,149 +One goal. Keep the work connected. + +2 +00:00:02,580 --> 00:00:04,080 +Delegate with context. + +3 +00:00:04,080 --> 00:00:06,185 +Keep corrections on the same task. + +4 +00:00:06,185 --> 00:00:07,580 +Let the owner adopt, + +5 +00:00:07,580 --> 00:00:09,587 +then return the artifact for review. + +6 +00:00:13,580 --> 00:00:16,221 +Goals, decisions, work and schedules. + +7 +00:00:17,080 --> 00:00:20,383 +Decisions, next actions, ongoing +observation. + +8 +00:00:20,580 --> 00:00:22,524 +Owners and waiting work, visible. + +9 +00:00:23,580 --> 00:00:26,436 +Trace questions, findings and plan +changes. + +10 +00:00:28,080 --> 00:00:30,120 +Inspect six process judgments. + +11 +00:00:30,120 --> 00:00:31,975 +Trace claims to the evidence. + +12 +00:00:31,975 --> 00:00:34,300 +Read measured results and costs +separately. + +13 +00:00:36,580 --> 00:00:37,930 +One solver record: + +14 +00:00:37,930 --> 00:00:40,270 +accuracy rose on duplicate graphs. + +15 +00:00:40,270 --> 00:00:41,749 +Control time rose too. + +16 +00:00:42,080 --> 00:00:43,920 +Keep failed paths and their reasons. + +17 +00:00:43,920 --> 00:00:46,246 +Use that evidence to choose the next step. + +18 +00:00:47,580 --> 00:00:50,097 +Check continuation, hosts and channels. + +19 +00:00:50,580 --> 00:00:51,890 +Your models and tools. + +20 +00:00:51,890 --> 00:00:53,823 +Durable work state with Loop X. + +21 +00:00:54,580 --> 00:00:57,580 +Loop X. Goals, work, evidence and +decisions. diff --git a/docs/assets/personal-workspace/loopx-research-workbench.en.vtt b/docs/assets/personal-workspace/loopx-research-workbench.en.vtt new file mode 100644 index 0000000000..9e9852cc77 --- /dev/null +++ b/docs/assets/personal-workspace/loopx-research-workbench.en.vtt @@ -0,0 +1,92 @@ +WEBVTT + +NOTE Narration captions for the unchanged 58-second film. +Sentence boundaries use measured pauses; no word-level alignment claim. + +intro-1 +00:00:00.080 --> 00:00:02.149 +One goal. Keep the work connected. + +steer-2 +00:00:02.580 --> 00:00:04.080 +Delegate with context. + +steer-3 +00:00:04.080 --> 00:00:06.185 +Keep corrections on the same task. + +steer-4 +00:00:06.185 --> 00:00:07.580 +Let the owner adopt, + +steer-5 +00:00:07.580 --> 00:00:09.587 +then return the artifact for review. + +kanban-6 +00:00:13.580 --> 00:00:16.221 +Goals, decisions, work and schedules. + +tasks-7 +00:00:17.080 --> 00:00:20.383 +Decisions, next actions, ongoing +observation. + +owners-8 +00:00:20.580 --> 00:00:22.524 +Owners and waiting work, visible. + +history-9 +00:00:23.580 --> 00:00:26.436 +Trace questions, findings and plan +changes. + +analysis-10 +00:00:28.080 --> 00:00:30.120 +Inspect six process judgments. + +analysis-11 +00:00:30.120 --> 00:00:31.975 +Trace claims to the evidence. + +analysis-12 +00:00:31.975 --> 00:00:34.300 +Read measured results and costs +separately. + +evidence-13 +00:00:36.580 --> 00:00:37.930 +One solver record: + +evidence-14 +00:00:37.930 --> 00:00:40.270 +accuracy rose on duplicate graphs. + +evidence-15 +00:00:40.270 --> 00:00:41.749 +Control time rose too. + +replan-16 +00:00:42.080 --> 00:00:43.920 +Keep failed paths and their reasons. + +replan-17 +00:00:43.920 --> 00:00:46.246 +Use that evidence to choose the next step. + +controls-18 +00:00:47.580 --> 00:00:50.097 +Check continuation, hosts and channels. + +hosts-19 +00:00:50.580 --> 00:00:51.890 +Your models and tools. + +hosts-20 +00:00:51.890 --> 00:00:53.823 +Durable work state with Loop X. + +outro-21 +00:00:54.580 --> 00:00:57.580 +Loop X. Goals, work, evidence and +decisions. diff --git a/docs/assets/personal-workspace/loopx-research-workbench.mp4 b/docs/assets/personal-workspace/loopx-research-workbench.mp4 new file mode 100644 index 0000000000..967637e832 Binary files /dev/null and b/docs/assets/personal-workspace/loopx-research-workbench.mp4 differ diff --git a/docs/assets/personal-workspace/loopx-research-workbench.transcript.md b/docs/assets/personal-workspace/loopx-research-workbench.transcript.md new file mode 100644 index 0000000000..6eebc275ff --- /dev/null +++ b/docs/assets/personal-workspace/loopx-research-workbench.transcript.md @@ -0,0 +1,110 @@ +# Research workbench film: transcript + +[Watch the 58-second film](https://github.com/user-attachments/assets/09883f57-7950-4bfa-bcfd-d80e836fb8ea) · +English subtitles: [VTT](loopx-research-workbench.en.vtt) / [SRT](loopx-research-workbench.en.srt) · +[简体中文](loopx-research-workbench.transcript.zh-CN.md) + +The English narration is reproduced below, with context for the on-screen +examples. Load a subtitle file alongside the [video](loopx-research-workbench.mp4) +in a player that supports external subtitles. The Chinese subtitle files +translate the narration; the film also contains Chinese on-screen labels. + +## 00:00–00:02.5 · Keep the work connected + +“One goal. Keep the work connected.” + +LoopX keeps goals, work, evidence and decisions durable. The Agent host provides +the model and tools. + +## 00:02.5–00:13.5 · Delegate with context + +“Delegate with context. Keep corrections on the same task. Let the owner adopt, +then return the artifact for review.” + +The illustrated brief carries the goal, constraints, correction, acceptance +criteria and return route. It finds an existing owner, who decides whether to +adopt the work. Nonconfidential App excerpts show the engineering handoff and +result return. Returning an artifact does not establish independent review, +merge or installed adoption. + +## 00:13.5–00:17 · See what needs you + +“Goals, decisions, work and schedules.” + +The Kanban view uses an authored workspace scenario to show work and attention +in one place. + +## 00:17–00:20.5 · Make the next action explicit + +“Decisions, next actions, ongoing observation.” + +The task view distinguishes work that can proceed, decisions that need a person +and observations that continue over time. + +## 00:20.5–00:23.5 · Keep responsibility visible + +“Owners and waiting work, visible.” + +Recorded task ownership shows who is responsible and which work is waiting. + +## 00:23.5–00:28 · Trace how a plan changed + +“Trace questions, findings and plan changes.” + +The recorded research history retains questions, findings and explicit changes +of direction, with the earlier state available for inspection. + +## 00:28–00:36.5 · Inspect the research + +“Inspect six process judgments. Trace claims to the evidence. Read measured +results and costs separately.” + +The packaged-UI example uses **synthetic data**. It shows six process judgments: +goal alignment, benefit testing, discrimination, information gain, evidence +adoption and avoidable repetition. Yes, no and unknown remain distinct. The +source-return path is an illustration of the reader's existing interaction. +These judgments are uncalibrated; they are separate from measured native +results and costs, and are not a total performance score. + +See the [trajectory assessment contract](../../reference/trajectory-assessment.md) +for definitions, eligibility and evidence requirements. + +## 00:36.5–00:42 · Read the scope and count the cost + +“One solver record: accuracy rose on duplicate graphs. Control time rose too.” + +The [historical solver record](https://github.com/loopx-project/loopx/blob/c9c3adf63f9771ef95f1fcca12b11ed7de1f69eb/apps/presentation/site/public/lab/time-machine/data/graph.json) +used two CPU threads and generated full-size joint-neighborhood graphs at 30% +partial duplication. Its recorded accuracy +rose from 41.2% to 85.0%. On the independent-edge control, accuracy stayed +unchanged while time increased by 32%. This is one historical record, not an +independent rerun or evidence of a causal LoopX-versus-baseline gain. + +## 00:42–00:47.5 · Use evidence in the next decision + +“Keep failed paths and their reasons. Use that evidence to choose the next step.” + +An illustrated workflow connects a failed path and its reason to the next +decision. This segment does not record live Agent activity. + +## 00:47.5–00:50.5 · Control continuation + +“Check continuation, hosts and channels.” + +The authored scenario shows continuation settings, execution hosts and +notification channels, including unconfigured and stopped states. + +## 00:50.5–00:54.5 · Bring your runtime + +“Your models and tools. Durable work state with Loop X.” + +The runtime supplies models and tools. LoopX retains work state across sessions. +This segment combines a mechanism illustration with a labelled Seedance 2.5 +concept animation. + +## 00:54.5–00:58 · LoopX + +“Loop X. Goals, work, evidence and decisions.” + +[Explore the workspace](../../../README.md#meet-the-personal-agent-workspace) · +[Get started](../../../README.md#try-loopx) diff --git a/docs/assets/personal-workspace/loopx-research-workbench.transcript.zh-CN.md b/docs/assets/personal-workspace/loopx-research-workbench.transcript.zh-CN.md new file mode 100644 index 0000000000..a989e103c1 --- /dev/null +++ b/docs/assets/personal-workspace/loopx-research-workbench.transcript.zh-CN.md @@ -0,0 +1,92 @@ +# 研发工作台影片：文字稿 + +[观看 58 秒影片](https://github.com/user-attachments/assets/09883f57-7950-4bfa-bcfd-d80e836fb8ea) · +中文旁白字幕：[VTT](loopx-research-workbench.zh-CN.vtt) / [SRT](loopx-research-workbench.zh-CN.srt) · +[English](loopx-research-workbench.transcript.md) + +以下是英文旁白的中文译文，并补充画面中示例的背景。可下载 +[影片](loopx-research-workbench.mp4)，在支持外挂字幕的播放器中载入字幕文件。 +中文字幕翻译旁白；影片中还保留了中文画面标签。 + +## 00:00–00:02.5 · 让工作衔接起来 + +“一个目标。让工作衔接起来。” + +LoopX 持久保留目标、工作、证据和决策。Agent 宿主提供模型与工具。 + +## 00:02.5–00:13.5 · 连同上下文一起委派 + +“连同上下文一起委派。纠正保留在同一项任务中。由负责人判断并接手，随后回传产物，供审查。” + +委派示意携带目标、约束、纠正、验收条件和回传路径，找到已有负责人，由接收方判断是否接手。 +非机密 App 片段展示工程任务的交接与结果回传。产物回传不代表已完成独立审查、合并或安装采用。 + +## 00:13.5–00:17 · 看清哪些事需要你 + +“目标、决策、工作和日程。” + +Kanban 使用编写的工作区示例，把工作和待关注事项放在一起。 + +## 00:17–00:20.5 · 明确下一步行动 + +“决策、下一步行动、持续观察。” + +任务视图区分可以继续执行的工作、需要人判断的事项和持续观察。 + +## 00:20.5–00:23.5 · 保留清晰的责任归属 + +“看清负责人和等待中的工作。” + +录制的任务归属画面展示由谁负责，以及哪些工作仍在等待。 + +## 00:23.5–00:28 · 追溯计划如何改变 + +“追溯问题、发现和计划变更。” + +录制的研究历史保留问题、发现和显式转向，也能回看之前的状态。 + +## 00:28–00:36.5 · 分析研究过程 + +“查看六类过程判断。沿着主张追溯证据。单独查看实测结果和成本。” + +打包 UI 示例使用**合成数据**，展示六类过程判断：目标对齐、收益检验、区分性、 +信息增益、证据采用和可避免重复。“是”“否”“未知”分别保留。 +引用回溯路径是对阅读器已有交互的示意。这些过程判断尚未校准，与原生实测结果、 +成本分开呈现，不合成为性能总分。 + +定义、适用范围和证据要求见[轨迹评估契约](../../reference/trajectory-assessment.md)。 + +## 00:36.5–00:42 · 核对条件，也计算成本 + +“一条求解记录：在重复图上，准确率上升。对照组的耗时也上升了。” + +这条[历史求解记录](https://github.com/loopx-project/loopx/blob/c9c3adf63f9771ef95f1fcca12b11ed7de1f69eb/apps/presentation/site/public/lab/time-machine/data/graph.json) +使用两个 CPU 线程，在完整尺寸的生成联合邻域图、30% 局部重复条件下，记录的准确率从 41.2% 上升到 85.0%。 +在独立边对照中，准确率未变，耗时增加 32%。这是一条历史记录，未经独立重跑， +不能据此得出 LoopX 相对基线的因果收益。 + +## 00:42–00:47.5 · 让证据进入下一次决策 + +“保留失败路线和原因。依据这些证据选择下一步。” + +工作流示意把失败路线及其原因连接到下一次决策；这一节没有录制实时 Agent 执行。 + +## 00:47.5–00:50.5 · 控制工作何时继续 + +“检查续行设置、执行端和渠道。” + +编写的场景展示续行设置、执行端和通知渠道，保留未配置与已停止状态。 + +## 00:50.5–00:54.5 · 使用你自己的运行时 + +“你自己的模型和工具。LoopX 保留持久的工作状态。” + +运行时提供模型和工具，LoopX 跨会话保留工作状态。 +这一节结合机制示意和已标注的 Seedance 2.5 概念动画。 + +## 00:54.5–00:58 · LoopX + +“LoopX：目标、工作、证据、判断。” + +[了解个人工作区](../../../README.zh-CN.md#认识个人-agent-工作区) · +[开始使用](../../../README.zh-CN.md#试用-loopx) diff --git a/docs/assets/personal-workspace/loopx-research-workbench.zh-CN.srt b/docs/assets/personal-workspace/loopx-research-workbench.zh-CN.srt new file mode 100644 index 0000000000..37ad641fc9 --- /dev/null +++ b/docs/assets/personal-workspace/loopx-research-workbench.zh-CN.srt @@ -0,0 +1,83 @@ +1 +00:00:00,080 --> 00:00:02,149 +一个目标。让工作衔接起来。 + +2 +00:00:02,580 --> 00:00:04,080 +连同上下文一起委派。 + +3 +00:00:04,080 --> 00:00:06,185 +纠正保留在同一项任务中。 + +4 +00:00:06,185 --> 00:00:07,580 +由负责人判断并接手， + +5 +00:00:07,580 --> 00:00:09,587 +随后回传产物，供审查。 + +6 +00:00:13,580 --> 00:00:16,221 +目标、决策、工作和日程。 + +7 +00:00:17,080 --> 00:00:20,383 +决策、下一步行动、持续观察。 + +8 +00:00:20,580 --> 00:00:22,524 +看清负责人和等待中的工作。 + +9 +00:00:23,580 --> 00:00:26,436 +追溯问题、发现和计划变更。 + +10 +00:00:28,080 --> 00:00:30,120 +查看六类过程判断。 + +11 +00:00:30,120 --> 00:00:31,975 +沿着主张追溯证据。 + +12 +00:00:31,975 --> 00:00:34,300 +单独查看实测结果和成本。 + +13 +00:00:36,580 --> 00:00:37,930 +一条求解记录： + +14 +00:00:37,930 --> 00:00:40,270 +在重复图上，准确率上升。 + +15 +00:00:40,270 --> 00:00:41,749 +对照组的耗时也上升了。 + +16 +00:00:42,080 --> 00:00:43,920 +保留失败路线和原因。 + +17 +00:00:43,920 --> 00:00:46,246 +依据这些证据选择下一步。 + +18 +00:00:47,580 --> 00:00:50,097 +检查续行设置、执行端和渠道。 + +19 +00:00:50,580 --> 00:00:51,890 +你自己的模型和工具。 + +20 +00:00:51,890 --> 00:00:53,823 +LoopX 保留持久的工作状态。 + +21 +00:00:54,580 --> 00:00:57,580 +LoopX：目标、工作、证据、判断。 diff --git a/docs/assets/personal-workspace/loopx-research-workbench.zh-CN.vtt b/docs/assets/personal-workspace/loopx-research-workbench.zh-CN.vtt new file mode 100644 index 0000000000..0da8d3fa08 --- /dev/null +++ b/docs/assets/personal-workspace/loopx-research-workbench.zh-CN.vtt @@ -0,0 +1,88 @@ +WEBVTT + +NOTE Narration captions for the unchanged 58-second film. +Sentence boundaries use measured pauses; no word-level alignment claim. + +intro-1 +00:00:00.080 --> 00:00:02.149 +一个目标。让工作衔接起来。 + +steer-2 +00:00:02.580 --> 00:00:04.080 +连同上下文一起委派。 + +steer-3 +00:00:04.080 --> 00:00:06.185 +纠正保留在同一项任务中。 + +steer-4 +00:00:06.185 --> 00:00:07.580 +由负责人判断并接手， + +steer-5 +00:00:07.580 --> 00:00:09.587 +随后回传产物，供审查。 + +kanban-6 +00:00:13.580 --> 00:00:16.221 +目标、决策、工作和日程。 + +tasks-7 +00:00:17.080 --> 00:00:20.383 +决策、下一步行动、持续观察。 + +owners-8 +00:00:20.580 --> 00:00:22.524 +看清负责人和等待中的工作。 + +history-9 +00:00:23.580 --> 00:00:26.436 +追溯问题、发现和计划变更。 + +analysis-10 +00:00:28.080 --> 00:00:30.120 +查看六类过程判断。 + +analysis-11 +00:00:30.120 --> 00:00:31.975 +沿着主张追溯证据。 + +analysis-12 +00:00:31.975 --> 00:00:34.300 +单独查看实测结果和成本。 + +evidence-13 +00:00:36.580 --> 00:00:37.930 +一条求解记录： + +evidence-14 +00:00:37.930 --> 00:00:40.270 +在重复图上，准确率上升。 + +evidence-15 +00:00:40.270 --> 00:00:41.749 +对照组的耗时也上升了。 + +replan-16 +00:00:42.080 --> 00:00:43.920 +保留失败路线和原因。 + +replan-17 +00:00:43.920 --> 00:00:46.246 +依据这些证据选择下一步。 + +controls-18 +00:00:47.580 --> 00:00:50.097 +检查续行设置、执行端和渠道。 + +hosts-19 +00:00:50.580 --> 00:00:51.890 +你自己的模型和工具。 + +hosts-20 +00:00:51.890 --> 00:00:53.823 +LoopX 保留持久的工作状态。 + +outro-21 +00:00:54.580 --> 00:00:57.580 +LoopX：目标、工作、证据、判断。 diff --git a/docs/assets/personal-workspace/research-workbench-hero.webp b/docs/assets/personal-workspace/research-workbench-hero.webp new file mode 100644 index 0000000000..3e1875f037 Binary files /dev/null and b/docs/assets/personal-workspace/research-workbench-hero.webp differ diff --git a/docs/assets/personal-workspace/research-workbench-hero.zh-CN.webp b/docs/assets/personal-workspace/research-workbench-hero.zh-CN.webp new file mode 100644 index 0000000000..b644635a18 Binary files /dev/null and b/docs/assets/personal-workspace/research-workbench-hero.zh-CN.webp differ diff --git a/docs/assets/personal-workspace/research-workbench-readme/film-a-distinct.png b/docs/assets/personal-workspace/research-workbench-readme/film-a-distinct.png new file mode 100644 index 0000000000..07a657cb22 Binary files /dev/null and b/docs/assets/personal-workspace/research-workbench-readme/film-a-distinct.png differ diff --git a/docs/assets/personal-workspace/research-workbench-readme/film-a-phone.png b/docs/assets/personal-workspace/research-workbench-readme/film-a-phone.png new file mode 100644 index 0000000000..b0aa352458 Binary files /dev/null and b/docs/assets/personal-workspace/research-workbench-readme/film-a-phone.png differ diff --git a/docs/assets/personal-workspace/research-workbench-readme/film-b-distinct.png b/docs/assets/personal-workspace/research-workbench-readme/film-b-distinct.png new file mode 100644 index 0000000000..fb178be7c1 Binary files /dev/null and b/docs/assets/personal-workspace/research-workbench-readme/film-b-distinct.png differ diff --git a/docs/assets/personal-workspace/research-workbench-readme/film-b-phone.png b/docs/assets/personal-workspace/research-workbench-readme/film-b-phone.png new file mode 100644 index 0000000000..c6bf22dd58 Binary files /dev/null and b/docs/assets/personal-workspace/research-workbench-readme/film-b-phone.png differ diff --git a/docs/assets/personal-workspace/research-workbench-readme/film-c-distinct.png b/docs/assets/personal-workspace/research-workbench-readme/film-c-distinct.png new file mode 100644 index 0000000000..7f1f72cb1a Binary files /dev/null and b/docs/assets/personal-workspace/research-workbench-readme/film-c-distinct.png differ diff --git a/docs/assets/personal-workspace/research-workbench-readme/film-c-phone.png b/docs/assets/personal-workspace/research-workbench-readme/film-c-phone.png new file mode 100644 index 0000000000..8949a1fd9e Binary files /dev/null and b/docs/assets/personal-workspace/research-workbench-readme/film-c-phone.png differ diff --git a/docs/assets/personal-workspace/research-workbench-readme/film-candidates-keyframes.png b/docs/assets/personal-workspace/research-workbench-readme/film-candidates-keyframes.png new file mode 100644 index 0000000000..7d1792a04f Binary files /dev/null and b/docs/assets/personal-workspace/research-workbench-readme/film-candidates-keyframes.png differ diff --git a/docs/assets/personal-workspace/research-workbench-readme/film-integrated-storyboard.png b/docs/assets/personal-workspace/research-workbench-readme/film-integrated-storyboard.png new file mode 100644 index 0000000000..53f3f617f4 Binary files /dev/null and b/docs/assets/personal-workspace/research-workbench-readme/film-integrated-storyboard.png differ diff --git a/docs/assets/personal-workspace/research-workbench-readme/film-opening-after.png b/docs/assets/personal-workspace/research-workbench-readme/film-opening-after.png new file mode 100644 index 0000000000..29b369634b Binary files /dev/null and b/docs/assets/personal-workspace/research-workbench-readme/film-opening-after.png differ diff --git a/docs/assets/personal-workspace/research-workbench-readme/film-opening-before.png b/docs/assets/personal-workspace/research-workbench-readme/film-opening-before.png new file mode 100644 index 0000000000..02fd5abd62 Binary files /dev/null and b/docs/assets/personal-workspace/research-workbench-readme/film-opening-before.png differ diff --git a/docs/assets/personal-workspace/research-workbench-readme/film-trajectory-analysis.png b/docs/assets/personal-workspace/research-workbench-readme/film-trajectory-analysis.png new file mode 100644 index 0000000000..0853ffa6f5 Binary files /dev/null and b/docs/assets/personal-workspace/research-workbench-readme/film-trajectory-analysis.png differ diff --git a/docs/assets/personal-workspace/research-workbench-readme/readme-en-after-desktop.png b/docs/assets/personal-workspace/research-workbench-readme/readme-en-after-desktop.png new file mode 100644 index 0000000000..8137a5dca6 Binary files /dev/null and b/docs/assets/personal-workspace/research-workbench-readme/readme-en-after-desktop.png differ diff --git a/docs/assets/personal-workspace/research-workbench-readme/readme-en-after-phone.png b/docs/assets/personal-workspace/research-workbench-readme/readme-en-after-phone.png new file mode 100644 index 0000000000..8a1e348f66 Binary files /dev/null and b/docs/assets/personal-workspace/research-workbench-readme/readme-en-after-phone.png differ diff --git a/docs/assets/personal-workspace/research-workbench-readme/readme-en-before-desktop.png b/docs/assets/personal-workspace/research-workbench-readme/readme-en-before-desktop.png new file mode 100644 index 0000000000..0be48e41b3 Binary files /dev/null and b/docs/assets/personal-workspace/research-workbench-readme/readme-en-before-desktop.png differ diff --git a/docs/assets/personal-workspace/research-workbench-readme/readme-en-before-phone.png b/docs/assets/personal-workspace/research-workbench-readme/readme-en-before-phone.png new file mode 100644 index 0000000000..b39ee6a740 Binary files /dev/null and b/docs/assets/personal-workspace/research-workbench-readme/readme-en-before-phone.png differ diff --git a/docs/assets/personal-workspace/research-workbench-readme/readme-zh-after-desktop.png b/docs/assets/personal-workspace/research-workbench-readme/readme-zh-after-desktop.png new file mode 100644 index 0000000000..2d38e03e1a Binary files /dev/null and b/docs/assets/personal-workspace/research-workbench-readme/readme-zh-after-desktop.png differ diff --git a/docs/assets/personal-workspace/research-workbench-readme/readme-zh-after-phone.png b/docs/assets/personal-workspace/research-workbench-readme/readme-zh-after-phone.png new file mode 100644 index 0000000000..b56c3fbb54 Binary files /dev/null and b/docs/assets/personal-workspace/research-workbench-readme/readme-zh-after-phone.png differ diff --git a/docs/assets/personal-workspace/research-workbench-readme/readme-zh-before-desktop.png b/docs/assets/personal-workspace/research-workbench-readme/readme-zh-before-desktop.png new file mode 100644 index 0000000000..caf57eaa72 Binary files /dev/null and b/docs/assets/personal-workspace/research-workbench-readme/readme-zh-before-desktop.png differ diff --git a/docs/assets/personal-workspace/research-workbench-readme/readme-zh-before-phone.png b/docs/assets/personal-workspace/research-workbench-readme/readme-zh-before-phone.png new file mode 100644 index 0000000000..a0b4ce94e0 Binary files /dev/null and b/docs/assets/personal-workspace/research-workbench-readme/readme-zh-before-phone.png differ diff --git a/docs/assets/personal-workspace/research-workbench-readme/transcript-en-desktop.png b/docs/assets/personal-workspace/research-workbench-readme/transcript-en-desktop.png new file mode 100644 index 0000000000..1e5d726dcc Binary files /dev/null and b/docs/assets/personal-workspace/research-workbench-readme/transcript-en-desktop.png differ diff --git a/docs/assets/personal-workspace/research-workbench-readme/transcript-zh-phone.png b/docs/assets/personal-workspace/research-workbench-readme/transcript-zh-phone.png new file mode 100644 index 0000000000..d01b807a14 Binary files /dev/null and b/docs/assets/personal-workspace/research-workbench-readme/transcript-zh-phone.png differ diff --git a/examples/public_entry/readme-demo-surface-smoke.py b/examples/public_entry/readme-demo-surface-smoke.py index d5792aaafa..8a86687ff4 100644 --- a/examples/public_entry/readme-demo-surface-smoke.py +++ b/examples/public_entry/readme-demo-surface-smoke.py @@ -33,7 +33,7 @@ def main() -> int: "long-horizon agents", "personal agent teams", "## Meet the Personal Agent Workspace", - "docs/assets/personal-workspace/workspace-hero.webp", + "docs/assets/personal-workspace/research-workbench-hero.webp", "docs/architecture/rfcs/capable-manager-semantic-handoff-v0.md", "## Why LoopX", "docs/assets/harness-neutral.en.svg", @@ -70,7 +70,7 @@ def main() -> int: assert required in readme, required for required in [ - "docs/assets/personal-workspace/workspace-hero.zh-CN.webp", + "docs/assets/personal-workspace/research-workbench-hero.zh-CN.webp", '', '', "docs/assets/harness-neutral.zh.svg",