WeChat AI Intelligence Board
Turning high-density WeChat discussions into a searchable intelligence system of quotes, links, products, and opportunities
Personal Intelligence System Case / 18
The WeChat AI Intelligence Board reorganizes fast-moving group discussions into a searchable, traceable, and reusable personal intelligence system.
The system filters high-value quotes, links, launches, methods, and demand signals from locally held chat records, then organizes them by topic, group, time, contributor, and relevance across overview, priority intelligence, product radar, link library, and contributor views. The deployed experience is protected and does not publicly expose chat content.
The documented snapshot covers 27 groups or conversations from 2025-12-13 to 2026-08-05.
Local processing parsed 125,882 messages into 6,527 high-value candidates, 4,852 deduplicated links, and 819 product leads.
The front end is static HTML/CSS/JS with PWA support, refreshed through local collection and rebuilding.
Cloudflare Worker provides protected access and asset serving; the raw chat database is not public.
The Job Is Not Archiving Chats but Reducing Judgment Cost
Group chats mix greetings, fragments, repetition, links, launches, and real needs. A plain archive merely moves the noise, so the first layer scores candidates through topics, length, links, launch signals, and methodological content.
The board retains quotes, contributor, group, time, links, and context so later product, content, and research decisions can return to evidence rather than detached summaries.
From Feed to Reusable Workbench
The overview reveals current themes, priority intelligence supports deep reading, product radar aggregates tools and launches, the link library preserves sharing context, and contributor views identify durable sources.
Private access, mobile adaptation, and PWA support make it a daily tool. Local collection remains separate from cloud serving so accounts, collection capability, and raw databases are not exposed publicly.
PROCESS
Key Decisions
Preserve quotes and context
Summaries support navigation while original expression supports verification and follow-up research.
Collect locally, serve privately
Collection requiring local sessions and file access stays local; the cloud serves only a protected static experience.
Model opportunity signals explicitly
Requests, unmet needs, willingness to pay, and recurring pain are separated from general discussion.
Role & Collaboration
I led focus themes, filtering logic, information architecture, module responsibilities, privacy boundaries, PWA experience, and deployment approach.
AI supported message structuring, candidate filtering, and development; I owned intelligence judgment, disclosure boundaries, and access control.
Validation, Outcome & Reflection
The deployed address is protected by email OTP and is not anonymously accessible, so this portfolio case provides no access button.
Counts come from the local 2026-08-05 build snapshot and will change with refreshes; they are not user-growth or commercial metrics.
A personal intelligence system should not pretend to read every chat for me; it should show what deserves attention and preserve a path back to evidence.
DEMO / 18
Product demo · 33 sec
A 720p private-dashboard demo covering the overview and primary navigation; raw chats, accounts, and local refresh workflows remain private.
