Product Updates

Thesora Ships Academic Research Module + Competitive Research

Thesora added an academic research report type, connecting to Crossref, OpenAlex, Unpaywall, arXiv, Semantic Scholar, PubMed, and DBLP; competitive research capability launched alongside it, with competitor matrices and feature/positioning benchmarking.

• Academic reviews cover five types: narrative, systematic, theoretical, technology, and software-engineering systematic literature reviews
• New journal profiling, article-type constraints, citation and full-text verification, and retraction checks — eligible types can run a deterministic PRISMA screening flow
• Competitive research covers competitor discovery, market-landscape assessment, comparison-dimension breakdown, and matrix generation, with evidence kept for every differentiating claim

Yanshi Ships Evolution Engine, Turning Failures into Engineering Rules

Yanshi added an evolution engine that attributes recurring failure patterns from signals — gate rejections, task errors, incidents, quality records, and user corrections — into rule-improvement proposals. Proposals only take effect after human approval, and subsequent tasks follow them automatically.

• Failure signals go through deterministic normalization, clustering, and attribution to surface recurring, evidence-backed failure patterns
• New-rule or rule-change proposals require human approval before entering the engineering rulebook; effectiveness can be reviewed by comparing recurrence and correction rates before/after
• Model backends now get a 9-dimension evaluation (tool use, structured output, instruction following, etc.), with adapted prompt suggestions for lower-consistency models
• New export of authorized, field-level de-identified training datasets, limited to super-admins and companies with explicit authorization

VeriQ Launches "MiaoBi" Content Generation — PPT, Docs, Reports, Charts

VeriQ added "MiaoBi" content generation, turning local materials and deep-research results into PPT decks, documents, in-depth reports, and charts, with support for anchoring multiple source documents and iterating across generation rounds.

• MiaoBi covers four scenarios — PPT decks, Word documents, in-depth reports, and charts — anchoring up to 5 documents/chapters at once
• Deep research can combine local material snippets with web results, keep citations, and hand results off to MiaoBi for continued generation
• App store presence keeps expanding, now live on the App Store, Tencent MyApp, Xiaomi, and other major channels

InsightForge Closes the Growth Loop: Image/Video Production + SEO + Competitor Tracking

Building on the short-drama studio, InsightForge added image and video production workbenches and SEO briefing, plus competitor public-channel tracking and complaint-signal analysis — closing the loop from content production to lead attribution and complaint feedback.

• Image workbench supports text-to-image and image-to-image; video workbench supports text-to-video and image-to-video, with generation parameters organized per target platform
• New SEO briefing and audit capability; competitor tracking collects public mentions, comparisons, and pain-point shifts by target and keyword
• Complaint signals are discovered across multiple public channels, then rule-prefiltered, structurally extracted, and semantically clustered into pain-point scores and action suggestions
• Publisher covers WeChat, Xiaohongshu, Zhihu, Weibo, Douyin, and Kuaishou, with high-risk content routed to human review

SuperTrade Adds Execution Observability, Expands Futures & Options Coverage

SuperTrade added an observable execution funnel — traceable from signal generation through approval, audit, tradability, and final settlement — and expanded futures to domestic and US markets with an options strategy library now spanning a dozen-plus spread and hedge structures. SuperTrade remains a personal, self-directed multi-asset research and trading-engineering exploration project and does not constitute investment advice.

• Execution center connects signal generation, approval entry, audit clearance, execution-threshold triggers, account tradability, order submission, and brokerage settlement end to end
• Order reconciliation is anchored to real brokerage settlement flow, compared against local order state, with discrepancies surfaced visually
• Futures gained daily bars, contract/instrument mapping, roll handling, and domestic open-interest, dragon-tiger list, and warehouse-receipt data; a 7-factor futures model shipped
• Options strategy library now covers protective puts, collars, calendar spreads, butterflies, straddles, iron condors, covered calls, and a dozen-plus other structures

BidScout WeChat Mini Program Launched

BidScout's government-procurement decision workbench launched a WeChat mini program, extending opportunity discovery, matching, and deadline reminders to mobile.

ShoeOps WeChat Mini Program Launched

ShoeOps launched WeChat mini programs for workers, owners, and finance staff, working alongside the H5 client to cover order intake, dispatch reporting, payroll, and reconciliation with real-time data sync across endpoints.

Thesora Goes Live in Production

Thesora, the AI deep-research platform for China's private markets, is officially live. Payment is now fully wired up, and the public landing page's self-serve signup flow is ready, with developers able to try research tools straight from the Playground.

• Payment flow upgraded for more reliable checkout, with fallback handling for failed transactions
• Public landing page shipped self-serve signup, in-page unlock, and one-click "generate a similar report" flow

VeriQ Knowledge Map v2 Launched

VeriQ shipped Knowledge Map v2 (chapter-structured + cognitive-level mind maps), a calendar-based learning goals system, and community "circles" — with quiz progress now unified across materials.

• Knowledge Map v2: batch generation, completeness checks, spam filtering, chapter ordering, and improved search
• Mind-map node question counts now aligned with quiz logic; empty nodes auto-trigger question generation; cognitive-level mind maps launched
• Learning tab gained a weekly-report card on top; cumulative quiz stats now unified across materials and profiles
• Community "circles" launched: discovery plaza, admin review, my-circles management, and cover uploads

Yanshi Open API + Mobile Remote Takeover

Yanshi launched an external open API (company-level key auth + multi-turn session interface + rate-limit protection) alongside mobile remote takeover of local Claude Code sessions.

• Open API launched: company-level key auth, multi-turn session interface, and rate-limit protection — ready to plug into your own systems
• Mobile app gained remote takeover of local Claude Code — scan a QR code to take over approvals and chat
• Artifact hub gained a draggable tree view and download permission controls

Thesora Multi-Hop Reasoning Engine + General Research Mode

Thesora completed its multi-hop reasoning engine and 3-tier evidence traceability chain, and shipped hard quality constraints for reports (A-grade evidence sourcing + multi-layer reasoning checks). It also promoted general research to a first-class report type — beyond private-market investment research, it now covers broader deep-research work like industry feasibility studies and competitive landscape analysis.

• General research promoted to a first-class report type: covers industry feasibility studies, competitive landscape analysis, and other broader deep-research work, not just financial investment research
• Hard report-quality constraints shipped: A-grade evidence sourcing + multi-layer reasoning checks, reducing plausible-but-wrong conclusions
• Evidence verification made more reliable, with automatic refunds for failed transactions
• Report library gained cross-report evidence search and clearer evidence presentation

Yanshi Notifications & Collaboration Polish

Desktop notifications for session/task completion are now clickable and trace back to their source; the artifact hub gained a Wiki page for accumulated project knowledge; a goal-delivery dashboard (spend/duration/model usage) went live.

• Desktop notifications for session/task completion fully wired up, clickable back to the originating session, task, or incident
• Artifact hub gained a Wiki page — project knowledge accumulates and is usable right after upload
• Goal-delivery dashboard launched, surfacing spend/duration/model usage at a glance

SuperTrade "Catch a Trade" Scanner Launched

SuperTrade launched its "catch a trade" panoramic stock-picking system, connecting signal scanning, one-click order placement, three-channel alert push, recommendation tracking, and post-trade review end to end — with real-time alerts for after-hours and pre-market momentum candidates.

• Panoramic view upgraded: persistent snapshots, aggregate view, per-symbol detail, and historical replay
• News-catalyst detection + sector co-movement + catalyst-driven opportunity tracks
• One-click order window + three-channel (email/SMS/in-app) alert push
• Recommendation tracking table + SELL-side signals + post-trade review cleanup; after-hours scanning opened up with real-time pre-market momentum alerts

SuperTrade Rebuilds Three-Market Reports

SuperTrade rebuilt its US/HK/A-share three-market intelligence reports, integrating Tushare A-share adjusted quotes, linking market reports with macro risk briefs, and moving narrative commentary to AI deep-dive analysis.

• Integrated a professional A-share market data source; strategy layer adapted per market
• US/HK/A-share three-market reports rebuilt to generate continuously
• Macro layer now linked with a global risk brief
• Narrative commentary moved to AI deep-dive analysis (thesis/theme logic/anomaly attribution), with rule-based templates as a fallback

Sypro Tech Website Launch

Official website launched, showcasing each product line's positioning, core capabilities, and technical methodology. Bilingual Chinese/English support with responsive design for desktop and mobile.

InsightForge Launches AI Short-Drama Studio

InsightForge added an AI short-drama production studio — multi-character dialogue dubbing, lip-sync and visual-consistency upgrades, cross-episode asset reuse via a series bible, and a topic-evaluation panel for pitching new series.

• Multi-character dialogue dubbing enabled by default, with synchronized lip-sync and visual-consistency upgrades
• Multi-episode series backend launched: series bible + cross-episode asset reuse + per-episode production gates
• Short-drama studio gained 5 core feature pages and a topic-evaluation panel

BidScout Launches In-Browser Bid Collaboration

BidScout integrated WPS WebOffice for in-browser collaborative bid editing, added supplier intelligence profiles (buyer tenure risk + competitor/lead-gen lookups), and connected an AI ingestion pipeline.

• In-browser collaborative bid editing via WPS WebOffice, supporting multi-person company-team collaboration
• New supplier intelligence profiles: buyer-tenure risk detection + competitor/lead-gen lookups
• AI ingestion connected for more timely opportunity discovery

VeriQ Demo Goes Live

VeriQ intelligent learning platform demo released publicly. Features document upload, AI auto-conversion to interactive learning materials, Bloom 6-level cognitive quizzing, knowledge graph visualization, and FSRS spaced repetition.

Yanshi 14-Stage Pipeline Connected

Yanshi AI autonomous development engine completed full 14-stage pipeline (discovery → design → code → test → review → merge). Git Worktree isolation and 5-layer budget protection passed stress testing.

ShoeOps Full Production Pipeline Connected

ShoeOps footwear production platform completed end-to-end digitalization from order intake → dispatch → reporting → reconciliation → payroll. AI dual-engine photo recognition (18 fields), 4 smart dispatch modes (keeping size runs intact), and piece-rate payroll auto-calculation with tamper-proof snapshots launched. Three-end coordination (admin / H5 / WeChat mini-program) with multi-tenant data isolation.

InsightForge Knowledge Distillation Pipeline Live

InsightForge completed 4-step AI knowledge distillation engine and content gap taxonomy framework (depth/angle/access). Multi-platform trend discovery and de-AI processor entered internal testing.

SpotAdvisor 21-Engine Integration

SpotAdvisor completed integration of 21 independent analysis engines covering business density, spending power, demographics, competition, and transportation accessibility. Natural language query interface and evidence-tracing map integration launched.

BidScout 5D Matching Engine Complete

BidScout procurement recommendation system completed 5D matching engine (business keywords, region, budget, procurement method, SME reservation) and 6 rigged-bid risk signal detection models. Multi-channel push pipeline launched.

SuperTrade 5D Fusion Engine Complete

SuperTrade quantitative trading system core architecture completed — 5D signal fusion engine (technical, sentiment, options flow, fundamental, quant factors) validated through internal backtesting. Bull-Bear adversarial debate mechanism launched.

SuperTrade Quant Prototype Kickoff

SuperTrade quantitative trading system kicked off, establishing the 5D signal fusion architecture (technical, sentiment, options flow, fundamental, quant factors) and completing data integration with the first backtesting framework prototype.

Sypro Tech Founded

Sypro Tech officially established, committing to the "AI-native × Vertical Industry" product strategy — first entering quantitative trading, content creation, commercial location, autonomous development, intelligent learning, government procurement, and footwear production management, and continually expanding.