zjz Lab

Simpler is better.

A project board for flagship launches, focused previews, and smaller supporting work, built with more space and less noise.

Project I

BlackHole Detector

一个用于捕捉异常信号的观测界面,把告警、置信度和原始轨迹压缩进同一块清晰预览,先看见结果,再进入判断。

Signal analysis Anomaly review Flagship build
BlackHole detector
Flag the unstable traces before they collapse into noise.
This surface ranks suspicious readings, keeps the raw telemetry visible, and lets the operator review anomalies without opening three different tools.
Detection layers
Gravity lens scan
Noise classifier
Event timeline
One surface, one decision path.
Inspect the next anomaly cluster

Project II

Microfiche Problem Extractor

这个项目把老档案、扫描页和破碎 OCR 文本里的问题点提取出来,最后整理成可以立刻行动的清单,而不是一堆散碎记录。

Archive parsing Issue extraction Research ops

Project III

Z Memory

一个把长期记忆、语义检索和上下文回看压进统一界面的记忆层,让“找回信息”像翻阅一页安静的系统面板。

Memory layer Context recall Knowledge retrieval
Z Memory
Memory retrieval should feel closer to reading than to debugging.
Z Memory surfaces related notes, prior decisions, and recall confidence in one place so the user can continue the task without opening a second layer.
Recall Thread history, reusable memory objects, and semantic matches.
Context Prior plans, attached files, and action summaries in one frame.
Goal Continue the work with less searching.

Project IV

Agentic Workflow

多代理协同最容易失控,所以这个项目重点不是炫技,而是把观察、拆解、执行和复盘排成一条可读、可追踪的流程线。

Multi-agent flow Reviewable steps Execution trace
Agentic workflow
Observe the task Capture the goal, constraints, and available context.
Assign the agents Split analysis, execution, and verification into clear roles.
Review the handoff Keep every transition visible instead of hiding it in logs.
Close the loop Summarize the outcome and keep the trace reusable.

Project V

Z Database

这个数据库项目关注的是“数据产品化”: 架构、关系、搜索和管理都被压成一个更轻、更直观的操作表面。

Data product Schema view Query surface
Z Database
Shape the schema into a quieter, usable surface.
Z Database turns relationships, table views, and filters into a product experience that reads clearly before the user writes a single query.
Data views
Schema explorer
Query surface
Relation graph
Inspect the selected collection

Project VI

Project Planner

一个更轻的规划工具,把目标、排期、负责人和提醒收进一条连续工作带里,不让计划页变成沉重的管理后台。

Planning board Timeline Execution view

Project VII

Agentic Trading

策略、信号、风控和执行反馈都在这里被压成更连贯的视图,让“交易代理”看起来像一条系统工作流,而不是一堆指标堆叠。

Trading system Signal engine Execution review
Agentic trading
One pane for strategy, signal strength, and execution status.
Agentic Trading keeps the model rationale readable, the execution risk visible, and the resulting position easy to review in one calm surface.
Strategy Hypothesis, entry logic, and market context side by side.
Risk Position size, exposure, and stop conditions without noise.
Loop Signal to execution to review.

Project VIII

Dynamic Resume

这个项目把简历从单一静态文档变成可重组表面,让内容跟岗位场景一起切换,而不是永远手动复制粘贴。

Resume builder Adaptive profile Role-specific export
Dynamic resume
Select the target role Choose product, research, design, or engineering emphasis.
Re-rank the projects Pull the most relevant work to the front automatically.
Rewrite the summary Adjust tone and emphasis without rewriting the whole document.
Export the final cut Keep the layout clean, fast, and role-specific.

Project IX

Neural Chat

一个更连续的对话界面,把上下文记忆、工具调用和多轮回复整合成同一条安静会话,不让用户在聊天和工作之间来回跳。

Conversation UI Context memory Tool-native chat
Neural chat
Summarize the thread, recall the brief, and draft the next move.
Neural Chat keeps memory, active tools, and live replies in one conversation surface so the next action feels like a continuation, not a context switch.
Context channels
Conversation memory
Attached tools
Live response state
Less interface, more continuity.
Continue the conversation...