This release makes the **AI Quant page more honest and easier to read**: activity rows follow real chronology, pending orders explain themselves, deferred signals stop looking like failures, follow cards stop blaming you for signals that were never dispatched, and attribution now shows both star trades and worst trades. 【AI Quant · Clearer Activity Details】 - Activity rows are sorted globally by time (newest first), consistent between the home preview and "View All" — no more old "deferred" rows floating above today's fills - Pending buys: when the current price is well above the target buy price, you'll see "wait for a pullback to the target" — explaining why it's still pending - Failed rows separately show when the signal was generated vs. when it was dispatched live, so you won't misread "still sending orders after the close" - Safely deferred after-hours signals are labeled neutrally as "deferred"; the red "execution failed" count only includes genuine failures worth your attention 【Follow Linkage · No More False Blame】 - Display-only signals that were never actually dispatched no longer show misleading "execution rate / execution deviation" metrics - New "View Strategy Pool" shortcut in the follow linkage section — jump straight to the strategy leaderboard 【Attribution · See Both Ends】 - Star trades now include open positions with unrealized gains, tagged "Holding" and distinct from closed trades - New "Worst Trades" section — compare winners and losers for a fuller review 【Smoother】 - Smoother scrolling on the AI Quant activity list for long histories — All AI content is objective, data-driven analysis for educational reference only and does not constitute investment advice.

This release makes the AI more honest and easier to read: see both the bull and bear cases (and each side's fatal flaws), get an honest heads-up before chasing a rally, spot crowded sectors at a glance, and finally see a mirror of your own trading discipline. 【Diagnosis · Bull vs. Bear Debate】 - No more single verdict: bull arguments, bear arguments, each side's fatal flaw, and the final ruling — laid out on one screen so you can judge for yourself 【Decision Card · Honest "Don't Chase" Alert】 - When the current price is well above the AI's entry zone, it tells you plainly: "chasing is risky, wait for a pullback" 【Strategy Lab · Sector Crowding Heatmap】 - See at a glance which sectors are crowded and over-extended — the redder, the more crowded 【Me · Trading Discipline Mirror】 - After a "don't chase" alert, did you hold back or pile in anyway? We track your "heed rate" and mirror your discipline back to you (shown once you have enough activity) — All AI content is objective, data-driven analysis for educational reference only and does not constitute investment advice.

This release is about **making points actually usable, a faster launch, and clearer live-trading protection**: points redemption now closes the loop — the points you've saved can be redeemed for perks, with a clear confirmation; cold start is noticeably faster; and when access lapses (e.g. Ultra expires), live auto-trading now pauses itself and guides you clearly, so nothing keeps running unattended. [Points You Can Actually Spend] - Points redemption loop is live: the entry now sits right where your points are shown, and saved points can be redeemed for perks instead of piling up unused - Successful redemption now shows a modal confirmation instead of flashing by; fixed an occasional "load failed" [Clearer Live-Trading Protection] - Lapse handling closed off: when access lapses (e.g. Ultra expires), live auto-trading pauses automatically to protect your funds, with an immediate heads-up - Clearer status wording: distinguishes "at-risk / action-needed" states, and the secondary button now reads "View Tutorial" with concrete next steps [Faster & More Stable] - Cold-start optimization: all non-first-frame initialization is deferred and AVAudioSession moved off the launch path — noticeably faster startup - Multiple live-trading and backend stability fixes

This release is about **making honesty the product**: AI self-positioning, membership tiers, and diagnosis track-records all shed the hype. The AI now calls itself a "research analysis assistant" instead of an investment advisor, membership lists are rewritten to reflect real capabilities, and Pro's core AI analysis becomes unlimited with no point deductions. The Diagnosis page adds a benchmark-adjusted "Overall AI Track Record" card plus a per-stock re-diagnosis log, and can generate a "Receipts" review poster in one tap. [Honesty & Compliance] - AI self-positioning unified as "research analysis assistant"; removed "investment advisor" and other license-implying wording across decision cards / rebalancing / insights / sim reviews - Pro / Ultra membership lists rewritten to real capabilities — dropped hollow "tiering / unlimited quota" perks, added back real ones like QMT live trading - Pro core AI analysis is now "unlimited (no point deduction)" — no more "100 points will be deducted" prompt; bonus / task points become redeemable for non-AI perks [More Transparent Diagnosis] - New "Overall AI Track Record · Benchmark-Adjusted" card: excess hit-rate at T+1 / T+5 / T+20, honestly stripping out the market's baseline move - The diagnosis detail page now has a permanent "This Stock's AI Re-Diagnosis Log": past calls + the actual moves that followed - One-tap "Receipts" review poster: bullish / bearish labels + factor insights + a signed disclaimer — owning our own calls - Review insights upgraded: real factor insights (consensus reversal / single-factor divergence) replace boilerplate [Discipline & Retention] - When your follow-trade buy price is meaningfully above the strategy's entry, a "chasing" nudge warns against blindly chasing a runaway - Follow-trade / sim return cards now note that sim fills favor pullback entries, so returns read optimistically — for reference only - New daily check-in "streak" flame on the My page, folded into the Task Center; Task Center trimmed to key tasks only - Push-permission request now explains the value first instead of popping up the moment you open the app [Live-Trading Fixes] - AI Quant "Current Holdings" now shows all positions beyond 10 (previously truncated at 10) - Operation history: after-hours deferred orders show a neutral "Deferred" instead of a misleading red "Failed" - Today's PnL backfills same-day realized gains from positions closed that day, matching broker reference PnL - Fixed phantom floating PnL where closed positions' stale unrealized PnL leaked into the live equity curve / holdings list - Intraday-T sell orders get a safety slice gate that never clears your full base position at once - Multiple backend stability fixes

This release is all about **trustworthy live AI-quant trading**: the live equity curve, operation history, and P&L attribution now all reflect real broker fills — no more sim numbers masquerading as real ones. The Watchlist / War Room gains a **Watchlist ⇄ Live** account switcher, and the AI Diagnosis poster swaps absolute price targets for risk-discipline metrics plus a per-stock re-diagnosis track record. [Live Data Now Reflects Reality] - Live equity curve no longer renders as sim: Y-axis amounts, crosshair, and starting capital now convert using your real live base capital (previously shown against the sim base) - Operation history backfills previously-missing real buys / sells, reconciled against real broker fills with dual-path dedup - P&L attribution now uses real broker fills with FIFO matching: win rate / trade count / realized P&L / actual commissions; pure-holding positions show "Holding · Open" instead of being misread as all losses - Fixed several sim/live mix-ups on the AI Quant page; live mode no longer renders sim amounts [Clearer Live Risk & Failure Reasons] - Human-readable failure reasons: a phantom order from a disconnected QMT now reads "Confirm QMT is logged in for trading on your PC" instead of the misleading "limit not reached, cancelled"; rejected-order reasons (e.g., price out of range) now pass through readably - Live risk copy changed to "per-order / daily buy cap", clarifying that sells (full stop-loss exits) are not capped - New ramp-up progressive-cap toggle with localized prompts; when off, only your per-order / daily caps apply - Follow-trade cards label the signal's base date; push wording states "base date + next-trading-day execution"; returns labeled "simulated" [Watchlist / War Room: Account Switcher] - The "Live only" filter becomes a segmented **Watchlist ⇄ Live** account switcher, moved above the overview card to emphasize switching accounts - Fixed the War Room header still showing sim "total assets" after switching to live (now shows real total assets / market value / available cash / floating P&L) - Fixed account-switch stutter and enlarged the switcher tap area (whole segment tappable) [AI Diagnosis Poster Redesign] - Absolute price targets replaced with risk discipline (percentage discipline + per-stock volatility / range-percentile scale) - Global win-rate replaced with a per-stock re-diagnosis track record (past calls + the actual moves that followed) - Recommendation label changed to "AI View · For Educational Reference" [Other] - New QMT premium-customization entry on the My page - Home fear-greed index no longer shows a stale snapshot; gauge restyled as a segmented arc slider - Risk page adds one-tap clearing of all alert rules - Factor-research daily card moved into the Strategy Lab; member stock-picks list adds one-tap add-to-watchlist - You can now edit others' strategies (auto-copied into your own private fork); "My Strategies" can publish to the leaderboard in one tap - Fixed Earnings Preview annual/quarterly selected button showing white-on-white in dark mode - Smoother crosshair dragging on the AI Quant equity curve - Multiple backend stability fixes

This release reworks the **Stock Detail page** and significantly improves overall smoothness. [Detail Page Redesign] - Reorganized into "Diagnosis / News / Overview" tabs for clearer navigation - On scroll, the price chart collapses into a mini price bar in the nav bar and the bottom tab bar hides — more reading space - Long-press the chart for a crosshair at any point, without blocking page scrolling - The "News" tab now auto-loads related news even before running a diagnosis [Overview Greatly Expanded] - Added company profile (valuation + snapshot + profitability), capital flow, analyst ratings & target price, and shareholder structure - Earnings Preview now charts actual vs. analyst consensus (annual) plus quarterly guidance - Added Hong Kong stock overview support [Smoother] - Fixed stutter when scrolling the detail page and when dragging the intraday-chart crosshair - Unified dark-mode visuals; nav bar no longer shows content bleeding through [Other] - New dedicated stock search entry on the Watchlist tab - Home screen supports Simple/Pro presets, module customization, and drag-to-reorder - Edit your custom strategies in-app and re-run the backtest - Fixed an issue where tapping an event push could fail to open the detail page

This release brings **One-Tap Posters**: five core scenarios now generate vertical share posters in one tap — AI Diagnosis, Strategy Backtest, Earnings Preview, Portfolio Battle Report, and AI Quant Battle Report. Scan to let friends see the full content + download the app. [Five Poster Scenarios] - AI Diagnosis: radar chart + decision card + recent trend + global win-rate - Strategy Backtest: equity curve + 4 key metrics + top trades - Earnings Preview: 5-quarter revenue/net profit + analyst consensus + ratings + target price (auto-backfill from income statement when express is sparse) - Portfolio Battle Report (Watchlist tab): real-position cumulative return + top winners/losers + equity curve - AI Quant Battle Report (AI Quant tab): paper-account win rate + closed trades + HS300 benchmark [Entry Design] - CTA reads "View" not "Share" — lowers psychological cost - AI Quant Battle Report uses indigo, visually distinct from orange share CTAs - Earnings Preview card sits below the risk panel on the stock detail page - Sheet expands fullscreen for browsing the full long image [Points Rewards] - First share: +200 points; daily share: +100 points - Auto-credited on WeChat callback, with toast feedback in the sheet [H5 Landing] - QR-scan auto-routes by scenario (?type=diagnosis/backtest/earnings/holdings) - Landing page leads with App download + 14-day premium membership trial [Other] - Privacy: nickname masked (Zhang**), portfolio amount hidden by default - Portfolio vs AI Quant Battle Reports use fully separate data (real positions vs paper account) - Multiple backend stability fixes

This release closes the **live-trading loop**: live mode restores the full PnL strip with today's PnL now including intraday holdings drift; three-tier live risk controls are hardened with a price-cage and an exchange limit-up/down hard gate stopping bad orders at the source; follow-trade cards now show a strategy-entry vs your-fill deviation so you never blindly chase a runaway. [Live-PnL Loop] - Live mode restores the three-column **Cumulative / Today / Holdings Float** strip - Today's PnL changed from "fees only" to **"fees + intraday holdings drift"** — a 530k account up 1% now correctly shows ~+5k - Actual commission now uses QMT broker-reported value, replacing the old inflated 5-yuan floor estimate - Live mode hides sim PnL fields to prevent the misleading "0%" badge [Three-Tier Live Risk] - Ramp-up daily cap scales by tier: **Aggressive ×2 / Balanced ×1.5 / Conservative unchanged** - Same-symbol same-direction filled signals are hard-deduped within the day - Aggressive bypass adds a **7-day sim-activity gate** - Limit orders >±2% from current are rejected, and the exchange limit-up/down gate is enforced — Aggressive is **not** exempt [Follow-Trade & Bridge] - Follow-trade card adds a **"Strategy entry X · You Y · Deviation Z%"** comparison - Bridge log shows **"Failed / Exchange-Rejected"** state; new **EOD reconciliation** (15:35 cron) finalizes stuck dispatches as cancelled / partial_filled - "Strategies dispatchable" renamed to **"Dispatch sources"**, count now includes AI-diagnosis dispatchers - QMT no-poll banner adds a restart hint; live-switch dialog reminds you to also flip the QMT-side run mode [AI Diagnosis & Backtest] - AI-diagnosis lean v2 adds a **bear-regime calibration**: in a bear market, BUY/OW is auto-demoted to HOLD on the card - Strategy-backtest chart switches to categorical x-axis with self-drawn axes, fixing tail-date alignment [Other] - Wallet upcoming cash-events fetch parallelized for faster load - Failed-order reasons no longer squeezed out by repeated English prefix concatenation - Multiple backend stability fixes

This release is centered on **Broker-Bridge v2**: risk preference now splits into three tiers — **Conservative / Balanced / Aggressive** — with the Aggressive tier becoming a true bypass that skips the 3% price-deviation gate and other systemic mis-blocks, fully aligned with sim dispatching. The bridge panel adds a **Simple / Detailed** view toggle, the new **Sim-Sync Mode** lets live accounts opt-in to also receive paper-strategy dispatches, and the new **Strategy Lab** drawer keeps your sim history and strategy-pool R&D data visible even after you switch to live. We also fixed a batch of false-alert banners around the lunch break and bridge start-up, plus a stale denominator in the watchlist daily-change calculation. [Three-Tier Risk Preference + Simple / Detailed View] - Risk preference expanded from "Conservative / Aggressive" to **Conservative / Balanced / Aggressive**; new accounts default to Balanced (the 3 existing Conservative accounts are left untouched) - Aggressive tier is a **true bypass**: skips the 3% limit-order price-deviation check (a limit order is already the user's own price guard — stacking a deviation check on top was the root cause of legitimate deep-pullback limit buys, e.g., fiona's, getting systemically blocked), skips price protection, and is fully aligned with sim dispatching. Only an admin-set disable will intercept - Bridge log now shows an **"Aggressive Bypass"** badge, so you can see at a glance which signals took the bypass path - Bridge panel adds a **Simple / Detailed** view toggle: Simple keeps only the main curve, key chips, and essential actions, hiding the subtitle line, watchlist mini-card, and redundant summary; Detailed keeps everything - Bridge subtitle unified to a single line: **"AI signals piped directly to Windows + QMT execution bridge"** — no more multi-line wraps

This release centers on **Sub-Segment Micro-Topics**: each of the 20 sub-segments inside an industry chain now opens its own dedicated 5-stage assembly-line comic, breaking down the segment's domestic-substitution progress across five stages — *Domestic Share / Overseas Benchmark / Domestic Leaders / Breakthrough Timing / Bottleneck* — so you can see at a glance where the segment is stuck and who is breaking through. We also shipped the long-awaited **broker-bridge auto-retry** for failed QMT signals, and split the opaque "Training: N" strategy counter into three transparent buckets so you can see exactly how close each strategy is to graduating. [Master One Industry Chain in a Day · Sub-Segment Micro-Topics (Cat-Quick-Read)] - Each of the 20 sub-segments in a top-level chain now has a "+" entry on its card, drilling into a dedicated 5-stage assembly-line comic breakdown - Each micro-topic includes its own **domestic-substitution progress card**: Domestic Share / Overseas Benchmark / Domestic Leaders / Breakthrough Timing / Bottleneck — five stages, single glance - Six core pillar companies (representative stocks) per segment, ranked by market cap + main-business purity, each tagged with role labels (Segment Leader / Import Substitution / Orders Materialized, etc.) - Visuals upgraded to scene-based industrial comics: crystal growth, cleanroom wafer fabs, optical-fiber welding, AOI inspection lines, data centers — no more "cat + arrow + circle" templates - New micro-topics auto-generated Monday / Wednesday / Friday on an event-driven schedule; several already live across humanoid-robotics, solid-state-battery, and other chains

This release rebuilds the representative-stock selection system for **"Master One Industry Chain in a Day"**, lifting accuracy from roughly 50% to 95%+. The CPO segment no longer fronts **Shijia Photonics** (real leaders are Innolight and Eoptolink); liquid-cooling no longer mixes in white-goods names; compute-leasing no longer surfaces healthcare or media stocks; large-model training now correctly highlights iFlytek instead of Great Wisdom. We also de-duplicate across segments so a mega-cap like Foxconn Industrial Internet only represents the segment most aligned with its core business, instead of polluting four or five segments at once.

Introducing an all-new column **"Master One Industry Chain in a Day"** — tap the featured card on the Discover tab to open a comic-style long infographic covering upstream / midstream / downstream plus four tiers with 20 sub-segments; each segment surfaces three representative stocks with positioning tags such as domestic substitution, scarcity, order realization, bottlenecks, leaders, and more. Native sharing to WeChat chats / Moments with rich preview cards. Fully automated on the backend: a new episode drops every Monday and Thursday at 09:00, drawn from a pool of 80+ rotating themes (AI compute, solid-state batteries, humanoid robots, commercial spaceflight…) ranked by popularity — roughly 40 weeks without repeats. [Master One Industry Chain in a Day · New Column] - The second card on Discover is the entry point: current episode cover (AI comic-style hero art), title, and subtitle - Dedicated topic screen: one hero image, four-color tiered grid (upstream core / midstream OEM / downstream applications / enabling layers), 20 AI hand-drawn segment icons, bottom industry-chain flow diagram, and same-day sentiment score - Each sub-segment lists three representative stocks (top 3 by market cap among Tonghuashun concept constituents), tagged with scarcity, domestic substitution, elasticity, bottleneck, order realization, leader, early mover, hidden champion, and similar roles - Horizontal "Past Episodes" strip at the bottom — browse every historical issue in one swipe - Fully automated: Monday / Thursday 09:00 drops; launch episode **Humanoid Robotics** with 20 segments (planetary roller screws, harmonic reducers, dexterous hands, AI brain, full robot bodies, compute training…) [Other Improvements & Fixes] - Industry-chain topic screen now uses system push navigation instead of a bottom sheet — swipe back from a standalone page for a more immersive flow - Fixed rare horizontal scrollbar on the right edge of the industry-chain H5 long graphic - Backend adds Alibaba Cloud OSS uploads for chain icons with public-read policy enabled - Debug builds validate API endpoint paths to prevent mistakes such as duplicated `/api` - Multiple backend performance optimizations and stability fixes

This release fully wires up Hong Kong financials in AI Diagnosis (now including earnings flash, improving freshness by 1–2 months), and fixes several caliber issues (HK on-hand cash, year-over-year back-derivation, gross margin, and more). Chart colors are now consistent with A-share convention (red = up, green = down). AI Quant Trading lets you customize per-order and daily caps, the bridge card surfaces strategy-capital exposure, and a series of decoding errors when running QMT simulation / back-test on Chinese Windows are fixed. Strategy following adds consecutive-loss protection — strategies on a losing streak get auto-downweighted or removed. [AI Quant Trading · Custom Risk Caps] - New "Risk Caps" section in Manage Config: edit per-order cap (up to ¥1,000,000) and daily cap (up to ¥5,000,000) directly in-app — no longer stuck on default ¥50,000 / ¥100,000 - The "Enable Auto Trade" switch now actually writes the backend safety policy — once on, AI signals go straight into the QMT queue without per-order confirmation; turning on live trading for the first time triggers a confirm dialog - Bridge status badges consolidated: "Live Running" and "Auto Enabled" now sit side-by-side on one row - QMT bridge card adds a "Strategy / Capital Exposure" panel: live strategy count, stage buckets (paper / canary live / production live), and today's dispatched count all at a glance - Order history cards show an execution-summary badge so you can see filled / partial / rejected / failed at a glance [QMT Bridge Stability · Multiple Fixes] - Fixed a GBK header decoding error when the QMT strategy runs simulation / back-test on Chinese Windows ("'gbk' codec can't decode byte") - Fixed collation conflicts in broker_bridge cross-table JOINs — errors in AI sim trading, reconciliation, and order sync are all resolved - AI sim trading now respects market holidays — May Day, New Year, etc. no longer generate phantom "executed" orders - Worker now has order / account sync channels — QMT-side positions and fills are pushed back to the app within 30 seconds, no manual refresh

Hotfix release: fixes a launch-time crash on iOS 18.0/18.0.1 introduced in v1.85 that prevented some users from opening the app even after a clean reinstall. Upgrading is strongly recommended. [Launch Crash Fix] - Fixed a startup crash on iOS 18.0/18.0.1 caused by initializing the audio engine before the audio session was activated (AVAudioEngine raised an NSException when connecting to mainMixerNode) - The streaming TTS player now defers audio-graph construction (attach/connect) until the user first taps "Listen" and after the audio session is active — no AVFAudio calls during launch - No feature impact: morning/closing briefing playback, voice diagnosis, and AI Investment Radio all continue to work as before

This release wires the AI Assistant's screener together with strategy generation: ask follow-ups directly from screener results and turn them into a strategy with one tap. Market labeling for full-market news is much more accurate now. The "AI Quant Trading" card also gets major improvements: heartbeat and bridge records auto-refresh, failure reasons are visible at a glance, and the Windows bridge setup is dramatically simpler. [AI Quant Trading · Card Improvements] - "Last Heartbeat" and "Recent Bridge Records" auto-refresh every 30 seconds and immediately on app foreground — no more switching tabs to see fresh data - Failed bridge records now show the actual reason (e.g. "QMT strategy not running", "bridge token expired"), instead of just a red "Failed" label - One-click Windows setup flow: unzip → double-click the installer → paste the strategy in QMT, finished in about 2 minutes - New Windows tray-mode operation: a status icon in the taskbar shows live online state and lets you switch paper/live anytime — no more keeping a black console window open [AI Assistant · Screener ↔ Strategy Generation] - Screener results now feed directly into AI Assistant follow-up chat, no longer short-circuited by the realtime-screener path - Promoted "Generate Strategy from Screener" entry — one tap pipes current conditions into the Strategy Lab for backtesting and tuning - Curated prompt "Refresh" works smoother; history conditions get a cleaner collapsed view - Keyboard-dismiss UX and result-area feedback polished [News & Market Labels] - Rebuilt news market classification: A-share ST/delisting/6-digit codes → A-shares, Fed/Nasdaq/Microsoft/Google → US, Hang Seng/HKEX → HK, preventing international geopolitical news from polluting A-share boards - HK/US single-stock news now fetched from real data sources with strict keyword filtering, replacing AI-fabricated content - US-name keywords broadened (Microsoft / Google / White House / North America trigger independently), improving HK/US single-stock classification accuracy [Discovery & Briefing] - New v1.84 WeChat-share briefing assets and Event-page copy/UX polish - Briefing-broadcast queue stability fix to avoid edge-case playback ordering issues - Discovery-page cache TTL jitter to smooth out peak traffic [Stability & Fixes] - Doubao TTS abort + cache throttling fix; TTS voice fields expanded - Alert historical-daily-K caching for faster response on repeat triggers - Chat-component copy unified across screens

This release introduces AI Stock Screener: describe industries, themes, market cap, fundamentals, valuation, moving averages, and volume conditions in natural language, and the app will screen the full market in real time with native result cards for Top Matches and Near Candidates. The AI Assistant also keeps its conversational screener flow, so you can keep asking follow-ups, adjust conditions, and compare candidates. [AI Stock Screener · Full-Market Search] - New "AI Stock Screener" entry on Home: full-market search, real-time quotes, broad evaluation, and match + candidate results - New native screener workspace with natural-language input, writing tips, example prompts, saved conditions, and a real-time result area - Example: "Semiconductors or communication equipment, market cap above 10B CNY, net profit growth >50%, near the 250-day moving average, recent volume surge" [Real-Time Match Pool · Easier to Read] - New structured "Real-Time Screener Match Pool" card: hit count, up-ratio, average/median gain, and market-cap distribution at a glance - Results now separate "Top Matches" from "Near Candidates", so fully matched stocks and close candidates are not mixed together - Each stock shows key evidence such as price change, market cap, PE, volume ratio, moving-average position, and financial label; tap to open stock detail [Natural-Language Understanding · More General Conditions] - Supports extracting market-cap ranges, industries/themes, earnings growth, valuation, moving averages, and volume/volume-ratio conditions from natural language - Fixed a case where full-market screeners were incorrectly handled as watchlist-only filtering - When users loosen or tighten market cap, growth rate, or industry scope, the execution criteria now follows the user's input instead of a fixed template [AI Assistant · Conversational Flow Preserved] - The AI Assistant's screener entry no longer jumps out to the workspace; it stays conversational for multi-turn follow-ups - You can keep asking questions like "why did these candidates miss", "run again with market cap up to 15B", or "only show stocks above the 60-day MA" - New-chat screener examples were refined to avoid mismatched examples or accidental watchlist-only filtering [Visual & Dark Mode Improvements] - Match and candidate cards now use clearer status differentiation: matches stand out more, candidates stay softer - Dark mode avoids large burgundy/brown surfaces, switching to low-saturation dark backgrounds with status-color accents - Mobile result-card layout improved to reduce cramped tables and text clipping

This is the "Trustworthy Research Loop" themed release. Every AI recommendation now shows its data evidence, gets one-tap actions (add to watchlist / set price alerts), auto-tracks 1/3/7-day outcomes, and lives in the new "My AI Track Record" on the My tab. Home gets a new "Today's 3 Things" card at the top. **See → Act → Follow-up** as a complete loop. [Evidence Layer · Data Cards] - New "Data Overview" collapsible section on AI Diagnosis / Multi-Model Panel / Daily Briefing: quote time, verification sources (Eastmoney + CLS + tushare cross-checked news), model confidence, missing-field hints, and 30-day direction-consistency rate, all visible at a glance - Confidence no longer uses red/green (v1.82 color bug fixed) — neutral state returns to the app's black-and-white tone - Data sources are dynamically filtered by actual fetch availability, no longer hardcoded; missing fundamentals / technicals are explicitly surfaced in "Missing Data" [Action Bridge · Next-Step Panel] - New "Next Step" panel at the bottom of the diagnosis report: one-tap Add-to-Watchlist + one-tap Set-Price-Alerts - Price alerts auto-generate 3 precise-price rules (breakout / breakdown) aligned with the AI decision card's entry / stop-loss / target levels - Supported on single-model diagnosis and multi-model panel [AI Track Record · Per-Action Review] - New "AI Track Record" entry card on the My tab: every AI action you've triggered is visible at a glance (watchlist / price alert) - Each record auto-links to the AI view at the time (bullish / bearish / neutral), trigger price, entry / stop / target levels - Every trading day at 15:35 CST, 1-day / 3-day / 7-day outcome is auto-filled, with 7-day direction correctness marked — for the first time you can truly see "how accurate is the AI" [Home · Today's 3 Things] - New "Today's 3 Things" card group at the top of Home (v1.83 flagship): one line each for Risk / Opportunity / Recap, so you know what to look at the moment you open the app - Risk row → tap to switch to the Risk Agent tab - Opportunity row → tap to open AI Diagnosis with auto-start - Recap row → yesterday's portfolio change summary [Daily Briefing Data Card] - The briefing player now shows a "Data Overview" collapsible below the playback controls, with the same semantics as diagnosis: quote time / verification sources / missing data - Shares the same UI component as diagnosis — consistent semantics across the app

This release centers on an iOS 18 launch-crash fix and a major upgrade to the AI Diagnosis "Quick Reference" — stop / entry / target now live on a structured price bar with a collapsible operations detail card, so buy/sell levels are readable at a glance. Daily Briefing and Watchlist also unify their day-PnL calculation, HK stocks get proper FX conversion, and low-liquidity briefings no longer fail. [Critical Stability Fix] - Fixed a launch crash on some iOS 18 devices (worked around AVAudioEngine rejecting Int16 non-interleaved formats) - StreamingTTSPlayer and Qwen realtime session stability hardened accordingly [AI Diagnosis · Structured Quick Reference] - New SVG price bar for Quick Reference: stop / entry / target rendered with three distinct markers (square / circle / triangle) and colors, current price highlighted separately - Paired price table plus collapsible "Operation Details" card makes reasoning and entry conditions readable at a glance - A-share colors back to "green down / red up"; dark-mode "current price" label switched to amber for clarity - Stop-loss zone moved to the upper-left of the bar, take-profit zone tightened — no more collisions with target or header - All price-table columns left-aligned, fixing an iOS WebKit issue where right/center alignment on headers was silently ignored - Report extractor schema split into stop / entry / target to reduce duplication and missed levels - Dynamic viewBox height prevents clipping on landscape or large-text accessibility settings [AI Diagnosis · Model Quality & Hallucination Guardrails] - Injects SW-industry sector breadth so the model distinguishes stock-α from sector-β more accurately - News cross-verification now pulls from two Tushare structured feeds (Eastmoney + CLS) to fact-check Qwen web-search output - Tiered evaluation (regime / conviction / bearish) is live — conclusions are no longer uniformly bullish - Real technical-analysis data is injected into diagnosis, eliminating fabricated price-pattern descriptions [Daily Briefing & Watchlist · Unified PnL] - Watchlist "Day PnL" now shows a percentage after the amount, matching position PnL and daily briefing - portfolioChangePct switched to "day PnL / yesterday's market value", fixing a case where +0.26% should have read +0.74% - Daily briefing todayPnl now performs HKD / CNY FX conversion, fixing an underestimate for HK holdings - FX data source switched from Eastmoney (blocked by Alibaba Cloud egress) to open.er-api.com — data once again stable - At HK market close, if "today's" data is not yet ready, the app no longer falls back to yesterday — no more stale post-market refreshes [Briefing · Low-Liquidity Coverage] - New A-share close snapshot refresh cron — STAR Market and thinly-traded names no longer trigger briefing_all_narratives_failed - HK close snapshot maintained by hk_daily every 5 minutes — low-liquidity HK names also stay briefable - Briefing-scenario cache TTL relaxed to 7200s, so even low-frequency tickers open instantly [Help Center · Getting-Started Entry] - A new "Quick Links" section at the top of Help Center — one tap to the iOS Getting-Started tutorial or the web help center - FAQ items rewritten to follow the real usage path, with entries like "What should I do when I first open the app?" and "How do I use AI Diagnosis and multi-model review?"

Big new feature: **AI Research Radio** — open it and you're tuned to your personal finance station, with morning, midday, and evening programs auto-lined up. CarPlay is supported too. Plus AI Assistant gains a "Deep Thinking" toggle, and stock diagnosis / multi-model panel / market briefings all get live-while-generating voice playback. [AI Research Radio · NEW] - New "Research Radio" entry in the Discover tab — tap in and it starts playing, no manual picking needed - Time-aware playlist: morning briefing / intraday limit-up alerts / post-market briefing / tomorrow's watchlist switch automatically — great for commuting, driving, or cooking - CarPlay support: a live finance radio in your car, with playback and program switching from lock screen / Dynamic Island - Alternating two-host dialogue (Longxiaochun + a male voice) and single-voice programs for a real-radio feel - Limit-up/limit-down flash bulletins (MVP): intraday limit hits trigger an automatic bulletin with a push notification for one-tap jump - Smooth host transitions between programs so back-to-back playback no longer feels abrupt - Lock-screen artwork now renders a dynamic black gradient with a white soundwave icon — no more gray square - Program sharing supported; a one-time tooltip on first visit points out where the radio entry lives [AI Assistant · Deep Thinking & Hallucination Guardrails] - New "Deep Thinking" toggle in full-screen AI Assistant — see the model's reasoning stream in real time, reasoning chain fully transparent - Home AI Q&A response time improved — no more 10–60 second blank wait - Data discipline upgrade: real-time quantitative questions hit hard data; qualitative business questions use judgment — fixes the "smarter = more hedging" problem - Four new structured intent cards to prevent fabrication: index PE/PB 10-year and 5-year historical percentiles, weekly market review, northbound-flow leaderboard, margin trading (market + stock), major-holder buying/selling - Basic / Pro / Ultra daily quotas shown by tier; exceeding the limit now surfaces an upgrade card instead of a blunt error

A full rollout of "AI investment voice" — the post-market briefing, morning briefing, and tomorrow preview are now wired into one **AI Investment Radio**, playing nonstop on your commute or in the car. Diagnosis reports, AI panels, and the market briefing all get a one-tap "Listen" button. Plus new account and privacy entries in Settings, with continued dark-mode polish. [AI Investment Radio · NEW · Pro/Ultra] - New "AI Investment Radio" card at the top of Discover — tap to enter an immersive player that auto-plays the full queue - Time-aware playlist: mornings lead with "Tomorrow Preview", market hours lead with "Morning Briefing", evenings lead with "Today's Briefing" — order adapts to when you listen - Floating MiniPlayer stays pinned across tabs with play / pause / close; closing also clears the lock-screen Now Playing entry - Lock screen & Bluetooth car: album-style artwork + progress + ±15 s scrub; works with headphone remotes, Bluetooth car audio, and phone cradles for hands-free commutes [Voice Share for Stock Analysis] - AI diagnosis reports add "Share voice version" — creates an H5 link with a built-in player - Recipients don't need the app: open in WeChat or a browser and the AI voice reads the analysis; the page links back to the App Store [One-Tap Listen · Streaming TTS] - "Listen" buttons added to AI diagnosis reports, AI panels, and the Home market briefing - Streaming synthesis — audio starts in about 0.5 s, no need to wait for the full text - Radio, briefing, TTS, and realtime voice share one audio coordinator: smooth hand-off, auto-pause on incoming calls [Ask AI · Realtime Voice Q&A] - "Ask AI · Xiaocan" entry inside the briefing player — speak a follow-up right after listening - AI answers in the briefing's context: conversational, under 30 characters, no re-reading [Settings] - Settings now has Account Info, Profile, and Privacy entries - Profile supports editing your nickname - Privacy toggles use the system default tint for a cleaner look [Pro Membership Value Report] - New "Your Pro Membership Value Report" card on My summarizes usage count, analysis credits saved, and key actions - Retention-aware: cancelled trial users also see a personalized report so they can review what they used before renewing - Ultra usage now takes the max of API and local sources for more accurate numbers [WeChat Community Entry] - New "Join User Group" entry on My — scan to join the official WeChat group - A one-time tooltip on first visit points out where the entry is - Long-press the QR to save to Photos; fixed an occasional crash on long-press [Stability Fixes] - Stock search no longer falls back to mock quotes — if there is no real quote, nothing fake is shown - Switching accounts now clears radio cache so the new account never hears the previous user's content; Basic users see a clear upgrade prompt instead of a silent 403 - Public share endpoints gain a 401 gate; share IDs use crypto-random + atomic writes to prevent enumeration and concurrent overwrites - Pre-market call-auction alerts no longer incorrectly show -100% phantom drops - Single-model diagnosis no longer gets stuck at "consolidating multi-source quotes" (fixes a diagnosis-worker loop crash) - Guest-to-login watchlist migration writes the tags column correctly - Financial data is no longer overridden by a mock; share-page JSON parsing is crash-safe - More stable watchlist loading and better cache hit rate for stock basic info - Trial-expiry push dedup extracted into its own module to cut down on duplicate sends

Membership and onboarding are more polished: trial progress is clearer, the membership card looks different for each tier, and AI Assistant colors are unified to the app's monochrome palette. [Trial Experience] - New countdown banner on Home shows remaining trial days and Pro features you haven't tried yet - New onboarding Step 5 "Try AI Rebalancing" appears after completing the first four steps - Usage stats displayed when daily diagnosis quota is reached [Membership Center Redesign] - My tab membership card now has distinct crown color, background, and border for Basic / Pro / Ultra - Expiry date moved below the tier name for a cleaner info hierarchy - Streamlined feature descriptions: Pro highlights Decision Card / Event Scoring / Portfolio Management; Ultra highlights QMT Auto-Trading and Deep Research [AI Assistant & War Room Polish] - AI Assistant selected states, icons, and send button unified to monochrome for a cleaner look - War Room adds local cache so reopening shows last data instantly instead of a blank screen [Account & Stability] - Binding a phone number after WeChat login now auto-merges existing accounts without data loss - Fixed share-page report IDs being incorrectly saved as invite codes - Watchlist position editing fixed, card gesture hints improved, auto-refresh after login

A dark-mode and top-bar polish release: contrast and layering in dark mode is meaningfully better, and the top of every tab now blends smoothly into the system status bar area. [Dark Mode Color Polish] - The four index chips (SSE / SZSE / ChiNext / STAR) inside the Home "AI Copilot" card now have clear layering against the parent card - Briefing body text now renders Chinese Markdown headings correctly (the ### markers used to leak through as literal characters) - Many previously near-invisible items are fixed: the "Pre-market" badge, the "AI Picks" sparkles icon, section header glyphs, and more - Fear & Greed gauge keeps its neutral range (41–60) in the app's black-and-white tone, reserving red/green only for extreme sentiment - 19 buttons and info tags across the app now have proper contrast: New Chat, Retry, Add Alert, Exit Plan type, and others no longer disappear into the background [Seamless Top Bar] - Home and My tabs: transparent top bar that scrolls with content, eliminating the color seam against the system status bar - Risk tab: sticky top bar with its blur material extending behind the status bar, sharing a single surface with the system area - Discover: "Today's Must-See" header is now tighter to the Dynamic Island - All hard dividers below the top bar are removed for a cleaner look [Other Improvements] - Tightened the visual gap between the top bar and the AI Copilot card on Home - Unfinished segments of the onboarding progress bar are visible in dark mode (previously almost invisible) - The new-user guide card now shares the design-system background, so it no longer looks off-tone next to neighboring cards

My Strategies is now live, event headlines can be tapped to read the full story, and intraday T-trading guardrail copy is easier to follow. [My Strategies] - New My Strategies card in the My tab shows your three most recent strategies with status and P&L preview - Full strategy list with pull-to-refresh and a clear backtest metric layout - Tap any Ready-state strategy to jump straight to its backtest detail view [Event Timeline Upgrade] - Event headlines are now tappable — a sheet opens with full article content, impact grade, and source - When article content is missing or too close to the title, it is automatically filled in on demand - Event radar now shows S / A / B grade events so high-impact items stand out [Intraday T-Trading] - "Risk-Reward Ratio" is renamed to "Profit-Loss Ratio" to better match trader terminology - Fixed conflicting copy when cost checks passed but the app still advised waiting — now shows "Cost structure is OK, waiting for price to enter the action range" - Tuned ratio thresholds and kept spread-based profit safeguards so smaller moves can still trigger cleanly [Other Improvements] - Home AI Copilot card is now fully tappable with a full-width "Enter Watchlist War Room" CTA - Strategy evaluation now shows a loading state instead of an empty "No backtest data" view - Backend multi-model analysis adds timeout protection and partial-cache quote completion for more stable analysis

Sim-trading attribution is now clearer, homepage indices update in real time, and backend analysis is noticeably faster with more accurate AI valuation data. [Attribution Upgrade] - Attribution results are now grouped by entry source and exit source, making signal origin and exit reasoning easy to follow - Sim-trading now deducts commissions for more realistic P&L - The position-reduction timing section is collapsed by default for a cleaner layout [Live Market Pulse on Home] - Four major indices (SSE / SZSE / ChiNext / STAR 50) auto-refresh every 4-6 seconds during trading hours - Fixed skeleton screen visibility in light mode for smoother loading transitions [Analysis Performance] - Batch quote queries dramatically reduce API call volume - Concurrent HK/US data fetching and cache reuse for faster analysis - Real-time PE/PB/market-cap snapshots injected into AI valuation prompts, eliminating occasional hallucinated figures

AI Assistant now feels more like a real investment panel. The new AI Group Discussion mode brings multiple AI perspectives into one guided session, while chat and intraday analysis visuals are smoother and easier to follow. [AI Group Discussion] - New Quick Ask / AI Group switch in AI Assistant, with stock, portfolio, and risk-event discussions - Each discussion starts by organizing recent public information and in-app data, then multiple AI roles respond in sequence - Final summary highlights consensus, disagreement, key risks, and what to verify next [AI Assistant Experience] - Group discussions now appear progressively for a more natural reading flow - Distinct avatars and identities make it clearer which assistant is speaking - Better moderator cues, freshness briefing, and summary copy make the experience feel more like a real panel [Intraday Analysis Visual Upgrade] - The scan animation in intraday analysis now aligns with the actual chart plot area - More stable waiting-state visuals make price action and key levels easier to track