How to use this site

screenshot — the real app

1234567891011

1. 1Stocks | Crypto scope
2. 2window
3. 3click a header to sort
4. 4Filters
5. 5Layouts
6. 6Columns
7. 7↓ CSV
8. 8type to filter tickers
9. 9status line
10. 10? help
11. 11🔔 bell

1. Pick your scope first. Stocks | Crypto at the top. Nothing carries across the wall, so if a column looks empty, check which side you are on.
2. Set the window (1h · 4h · 1d · 7d · 30d). Every column that says "over the window" moves together: vs S&P, vs tape, ΔOI, Squeeze, Carry, Avg Range, RVOL. Use 1d for the day, 7d for the week's story, 30d for positioning that has had time to build.
3. Sort by clicking a header. Click again to flip. The most useful sorts are not price changes: try vs tape (who is strong on a weak day), RVOL (who is being traded unusually), ΔOI (where money is arriving) and Squeeze.
4. Filter the noise. Filters holds volume and open-interest floors (in millions; 500k and 2.5b work too). A floor of a few million in 24h volume removes the names you could not exit at size.
5. Show the columns you actually use. Columns shows, hides and drag-reorders. Several useful ones are hidden by default: the moving averages, VWAP, the intraday opens, DownCap.
6. Hover before you conclude. Every cell tells you its inputs, its window and its sample size. A shaded cell scales with the size of the move, so a wall of deep red 7d cells is the breadth read.
7. Export with ↓ CSV — you get exactly the rows, columns, sort and filters on screen.

screenshot — the real app

123456

1. 1volume floor (millions)
2. 2OI floor
3. 3link a wallet (public address)
4. 4★ only
5. 5◢ noted
6. 6⬡ held

screenshot — the real app

12

1. 1sorted by vs tape
2. 2★ star = watchlist

• group → sectors / industries. The same table with one row per sector. Each cell is the weighted average of the rows you are showing — and Best · Worst tells you whether the sector number is everyone or one name dragging the rest. Click a group row to drill into its members.
• The action strip under the table (click ▸ Action). HEATING is what is starting: acceleration confirmed by flow. COOLING is what is still ahead of the tape but has stopped moving — with a high momentum score that is the take-profit read. STRONGEST BID is who got bought on the dip in the last four hours.

screenshot — the real app

1

1. 1the drawer

1. Open it by clicking a row, pressing Enter on a highlighted row, or Ctrl/⌘+K and typing the ticker.
2. Read it top to bottom. Header and event risk first (is there an earnings print or a macro event inside your horizon?), then where the 30d return happened (off-hours vs cash session), then the metrics grid, then candles and the OI / funding history, then news and the signal record.
3. Use the focus chip. Opening a name sets the ◎ focused ticker. Now Trend, Charts, Report and Correlation open on that same name — click the chip in the status line to reopen the drawer from anywhere.
4. Write while you are there. The Notes panel sits above the charts. That is on purpose: everything else in the drawer is the server's read; the note is yours.

• The 30d split is the underrated panel. A name that made all its return overnight is a different hold from one that made it during the session — and the funding you pay to hold overnight is right beside it.
• Co-movers and hedges are computed on 90 days of daily returns. Before you add a name, check whether it is a co-mover of something you already hold.
• The signal record is the out-of-sample history of every claim the engine made on this name. A name with a poor record is a name where the engine's flags deserve less weight.
• Pre-earnings setup. A name reporting within five sessions gets its setup card in the drawer — reaction study, positioning into the print, run-up, implied vs typical, and one verdict line. See Calendar.
• Every section header has a ⤴. Share the metrics, the reaction study, a sparkline or the candles straight into a conversation.
• The URL follows you. #t=NVDA in the address bar is a shareable link straight to this drawer.

1. Star names with the ★ beside the ticker. Starred names float to the top of the table; ★ only in Filters shows nothing else. The watchlist follows your account, so your phone has it too.
2. Arrange the table once — columns, order, sort, window, volume/OI floors, ★-only.
3. Layouts → save, and give it a name that says what it is for: positioning, open, swing.
4. Switch layouts from the same menu. A • beside the active name means you have drifted from what you saved; save again or switch back.

• Make a layout per question, not per day. A "positioning" layout with Funding, ΔOI, Squeeze, OI/Vol and Carry; a "strength" layout with vs tape, RVOL, DownCap and Momentum. Switching layouts is faster than re-sorting.
• Saved layouts sync; the active one does not. Your phone can run its own layout while the desktop runs another — that is deliberate, not a bug.
• New columns are never dropped from an old layout. When the app adds a column, your saved view gains it (hidden if it is hidden by default) rather than silently losing it.
• The ticker search box is not part of a layout — it is meant to be typed and cleared.

screenshot — the real app

1

1. 1Layouts menu

screenshot — the real app

1

1. 1show / hide / drag to reorder

1. Pick a side — the long board or the short board.
2. Read the score. Each name gets a rung per timeframe (daily, 12h, 4h, 1h); green means price is above a stacked EMA 13/21 ribbon on that rung. 4/4 is an established trend on every timeframe; 2–3/4 usually means the higher timeframes lead and the lower ones are pulling back.
3. Look for RETEST. The badge means price just probed into the ribbon while the close held the right side — the continuation pullback. The volume multiple beside it tells you whether the level is being fought (≥2×) or quietly respected (~1×).
4. Check Width and Age. Two 4/4 trends are not the same trade when one rides a hair-thin ribbon and the other a wide one. Age is how many days the daily rung has been stacked — a day-3 trend is an entry, a day-40 trend is a chase. The board ranks fresh-first for exactly that reason.
5. Click the chart button on a row to see the rung with the ribbon and the retest zone drawn at the ladder's own levels.

screenshot — the real app

1

1. 1long / short side

• Δ21 is the live distance from the hourly EMA21 — your proximity-to-entry number.
• The badge you see is the same condition that becomes a signal in the ledger. The board and the signal cannot disagree.
• A name missing a rung (too little history) is excluded, not guessed — the header counts them.

1. GRID puts up to eight names side by side. Choose panes (4, 6 or 8), then type into add ticker… to fill them; each pane keeps its own timeframe.
2. MTF WALL puts one name across timeframes — the same day seen at 15m, 1h, 4h and 1d (more rungs with more panes).
3. LINK shares the crosshair and zoom across panes (always on in the wall), so hovering one bar lines up the same moment everywhere. RESET returns every pane to its default window.
4. EMA overlays two moving averages of your choosing on every pane; SAVE keeps the whole arrangement as a named chart layout for this browser.

• Short timeframes read the local 5-minute archive; 4h and above read the deep daily archive. Each pane says which — they are never blended in one chart.
• Warm-up bars for an EMA stay empty rather than plotting an unconverged line.

1. Choose size — 24h volume or open interest. Volume shows where the activity is; OI shows where the positioning is.
2. Choose the window — colour is the move over it.
3. Click a tile to open the name.

• Size by OI and window 7d is the fastest way to see where a week of red actually hurt.

screenshot — the real app

1234

1. 1GRID or MTF WALL
2. 2panes
3. 3add ticker…
4. 4LINK

screenshot — the real app

1. Start on the flow map. Each bubble is a sector. Right = money flowing in with conviction (price plus open interest); up = activity (volume plus volatility); size = 24h volume. Top-right is accumulation, top-left is distribution, the bottom half is quiet.
2. Flip to Leaders to see who is beating the benchmark (right) and whether the lead is growing (up) or fading. "Catching up" is where early rotation shows; "Cooling" is where leadership goes to die.
3. Read Cohesion on the rotation board. High = the label trades as one thing, so any member will do; low = a stock-picker's sector where the name matters more than the theme.
4. Click a bubble for its members with their own contribution to the composite.
5. Try equal-weight. By volume, a mega-cap can be the whole sector's story; equal-weight shows whether the rest agree.

screenshot — the real app

123

1. 1flow map / leaders
2. 2window
3. 3weighting

• The sector × sector correlation matrix catches one risk factor wearing two names — Semis and Software moving together is one bet, not two.
• Grouping switches from GICS sectors to curated industries (Semis vs Software vs Memory). Industries are display-only; the signal engine keeps pooling on the eleven sectors.

1. Matrix. Names are cluster-ordered, so the blocks along the diagonal are real co-movement families. Hover a cell for the correlation and the number of days behind it — a ±0.9 on twelve days is noise. Use the focus search to isolate a few names; widen or narrow with top by vol.
2. Click a cell for the pair view: the hedge ratio (β, units of B per unit of A), the spread between them with its mean and ±1σ band, and the z-score. Beyond ±1.5–2 one side is rich or cheap against the other. The rolling correlation underneath says whether the relationship is stable enough to trust.
3. Click a ticker label for its co-movers (proxies and contagion) on one side and its natural hedges on the other.
4. COMP/G overlays several names rebased to 100 at an anchor you choose — presets, a date, or drag the amber line. Spread mode shows each name minus the basket of the visible names.
5. Ratio draws synthetic pair candles (A divided by B) with a properly honest EMA200 — drawn only where 200 bars exist. BTC may face a stock here, aligned on the hours both traded.

• Type comp AAPL MSFT NVDA or ratio NVDA/BTC 4h into the Ask console to jump straight to these views.

1. Read the colours. Red = longs pay (the crowd is long; holding costs money). Green = longs receive. Every row is also labelled with its signed number, so you never depend on colour alone.
2. Pick a resolution — 1h, 8h or 24h. In the default annualized unit, the timeframe is only a zoom: one market reads one number on every grid.
3. Compare now against the mean. Each row ends with its window mean and its current funding in the same unit. Current above mean = carry building; below = fading.
4. Sort by carry side, by |carry|, on the mean or on the now column.
5. Switch to per-bucket only when you want the actual amount a 1× long paid over a bucket — that unit changes with the timeframe (about 8× larger per 8h than per 1h).

• Hatched cells are honest gaps — the data covered under half the bucket's hours. They are never drawn as "flat", because a gap and a genuinely flat market mean opposite things to anyone sizing a position.
• The newest column is always the last complete bucket.
• Rows rank by open interest; show top 25, 50 or everything.

1. Start with the decomposition. Three equity curves: what an overnight hold (close→open), a weekend hold and a cash-session hold each paid, pooled across the equity universe.
2. Watch the shaded band. That is the running funding drag. An edge that dies inside the band is a donation.
3. Read the clocks — hour-of-day activity in each market's own norm, the funding clock, the day-of-week heatmap — with sample-size humility; hover for the breadth behind every point.

screenshot — the real app

• A rising overnight curve against a flat cash curve is the classic overnight effect. The drawer's where the 30d return happened panel is the same question asked of one name.
• Panels that need more history than exists yet stay empty rather than showing thin data.
• None of this exists in crypto scope — a 24/7 tape has no sessions.

1. Line 1 — the event, a plain-language reading, when it appeared, and the claim it is scored against (the frozen mark and the time to resolution).
2. Line 2 — the evidence, always in the same order: scope · n · median outcome · hit rate · expectancy · horizon. n is the number that matters. A pill saying unproven means no history; neg exp means past occurrences lost money on average.
3. Line 3 — the mechanical play: side, target, the level that voids it, reward:risk. These are frozen when the claim opens and never move.
4. Line 4 — the one corroborating thing to watch.

screenshot — the real app

1. Sort by score, but read n first. Score is unusualness now plus historical edge; a high score on n=3 is a story, not a base rate.
2. Prefer ★ prime — ≥60% hit, positive expectancy and sound structure at fire time.
3. Mind the decay. A signal past its claimed horizon dims (amber) and drops at twice the horizon. A stale signal is not a signal.
4. Check the earnings guard. A card flagged for a print inside the horizon has its evidence capped, on purpose, and says so.
5. Search a ticker to see its complete claim-by-claim history — what fired, at what mark, what it claimed, what happened. Do this before you trust any card on that name.
6. Look at the accuracy panel occasionally: live hit vs claimed, the profit factor (which catches the 55%-hit signal that still loses money), and the calibration buckets that audit the ranking itself.

• ⇄ on a card means long and short fired on the same name — flagged, and nobody gets a confluence bonus.
• Confluence (several same-side conditions on one name) only earns a bonus once the record proves agreement helps. It is measured, not assumed.
• The Actionable taboperator-only by default is the subset of these claims that are at their trigger right now, with reward:risk net of funding.
• The Backtest taboperator-only by default tests a ranking rule on the whole universe — or, with a name in its target box (or ★ watchlist), asks the smaller question "does this signal work on NVDA?". Under 10 round trips it tells you the Sharpe is an anecdote. Its D1 retest study replays the Trend board's RETEST over every closed day and reports the edge over simply being in the trend.

1. Read the groups top to bottom. Touching now is a candle still forming on a line (soonest close first) — it has not resolved. Fresh resolved within the last 3 bars (4H) or 2 sessions (1D): that is the actionable window. Fading is dimmed context, not a setup.
2. Read the title as the verdict. Testing support / Testing resistance while the candle forms; at its close it becomes Support held, Resistance held, Broke down or Reclaimed.
3. Use the action line as your stop. Every card names the level and the close that invalidates it — "Bullish while above 148.05. A 4H close below it invalidates."
4. Watch the outcome. ✓ worked means price has gone at least 1σ the expected way; ✕ failed means a later close went back through the line, and the card fades at once.
5. Look ahead with Near a line. Every line within 1.5σ of the mark, nearest first, with the % beside it. stacked means the 50 and the 200 sit within half a sigma of each other — one touch tests both.
6. Narrow it. The controls row: TF (both · 4H · 1D), EMA (both · 50 · 200), View (by stage · newest first) and Show (all · live + fresh).

• Distances are in σ, not %. σ is the rung's own bar-return spread, so 0.5σ is equally close on BTC and on gold. The % is beside it for when you want it.
• NEW marks touches since your last visit (kept per browser; the first visit marks nothing).
• One card per episode. After a card the line has to re-arm — a close at least 1σ clear — before it can open another; touches before that fold into the same card as retouches.
• 1D is the US session on a stock or index: weekend perp prints fold into Monday's candle, so the tab, the chart and the alert agree on what a daily candle is.
• The bell has matching alert classes — ma200, ma50 (close-confirmed) and touch200 / touch50 (intrabar, opt-in) — operator-only while the tab soaks.

1. Upcoming — the next two weeks of earnings for the universe, grouped by day with the session: BMO before the open, AMC after the close. Starred names float to the top of each day.
2. Reported — the last two days stay on the tab with actual vs estimate, beat/miss, and the print's own reaction candle. The tape's verdict is the move beside the EPS verdict; they often disagree.
3. Reaction study — for each name: how many prints, the average and median move the next session, how often up vs down, gap behaviour, and the move as a multiple of the name's usual daily range. That last one is the sizing number.
4. Macro rows — FOMC, CPI, payrolls, PPI, retail sales, GDP, PCE, with the prior print shown ahead of the release and prior → actual after.
5. Setups — the collapsible strip at the top: a pre-earnings setup card for every name reporting within the next 5 US sessions. Each card puts four things side by side: the reaction study (typical move, up/down split, gap-and-hold vs gap-and-fade), positioning into the print (funding and its percentile against the name's own month, OI change over the run-up, premium), the run-up against how this name usually drifts before a print, and implied vs typical (is volatility compressed or already elevated?). A verdict line sums it up from fixed rules — crowded long into print (funding p92, OI +18% in 5 sessions); typical move ±6.1%, gaps and fades 7/10. No AI writes it.

• You do not have to open the tab to be warned: the E badge on the Markets ticker cell (solid = today, hollow = tomorrow) and the macro banner under the nav read the same calendar.
• Indices, FX, commodities and baskets never report — the coverage line says how many names are eligible so absence is not a mystery.
• The same setup card appears as Pre-earnings setup in the ticker drawer of any name that has one. A block the server cannot fill says n/a and why (no study yet, funding history too short…) rather than showing a guess.
• A setup card is a read of who is positioned, never a call: crowded long says who is exposed, not who is right.

1. Filter by ticker with /. The drawer's news panel is the same feed, already filtered to the open name.
2. Read the badges. Amber = earnings within a week; red = an open signal on that name; FL = an SEC filing, with material forms marked.
3. The window is 72 hours. Older is history, not news.

• Type news nvda in the Ask console for the same slice as a table.
• Material filings can also reach you as an alert class — turn it on in the bell.

1. Open it from the drawer's AI report →, the Report tab with a focused ticker, or Shift+Enter in the command palette.
2. Read the picture and the flags first, then the scenarios. The colour follows the money: an adverse scenario renders red even if the model called it a "target".
3. Weigh the action block the way you weigh a card: its EV-at-entry is computed server-side, and a plan that prices negative is downgraded to wait before it reaches you.
4. Check the track record at the bottom — every directional read is ledgered and graded out of sample like any signal.
5. Regenerate only when something material changed; the button unlocks on a cooldown or on a moved mark, a fresh signal or a print.

• Reports are shared: the first person to generate one caches it for everyone. Generation is capped per person and per group per day, so do not spend it on a name nothing happened to.
• Sector and basket reports exist too: report sector Semis, report basket AAPL MSFT NVDA.
• It is a reading of numbers other tabs already show. If it cites something you cannot find on the board, that is a reason to trust it less, not more.

screenshot — the real app

1. Open the drawer on a name and click ＋ note in the Notes panel above the charts.
2. Write what you actually think — a level, a reason, a rule for yourself. One honest line beats a paragraph of hedging.
3. Tag it with #tags in the body: #thesis, #level, #mistake. Tags are read out of the text, so you never have to manage them.
4. Save. The note is stamped with the mark and the time, once.
5. Find it again on the Notes taboperator-only by default — newest first, searchable over bodies and tickers, filterable by tag — or by the ◢ post-it in the Markets ticker cell (bright within a week, dimmer to a month, hollow after). ◢ noted in Filters narrows the table to names you have written on.

• Editing changes the words and nothing else. The timestamp and the price stand, and the edit is disclosed. If your view changed, write a new note — that is the record you will want.
• A name rotated out of the universe keeps its note, greyed and labelled.
• Writing needs the operator to have turned on note-writing for you; reading the tab never grants the pen.
• A call in Messages can be written into the notes book with one click, keeping its stamp.

screenshot — the real app

1

1. 1calls ↗ — opens the Calls tab

1. Pick a conversation in the rail: a one-to-one with a member, a group, or a topic board anyone can join. Groups are created from the rail with a name; topic boards are standing posts for an idea, a ticker thesis, a theme.
2. Name tickers with a dollar sign. $HOOD is stamped with the price at the moment you send it, and every later read shows the move since. Write short $HOOD and it scores as a short; long $HOOD or buy $HOOD as a long. Calls are public — the next section, Making a call, is worth reading before your first one.
3. Post a table instead of describing one. /top funding 5, /nvda, /screen rvol>2, /earnings today — the result lands in the thread under your name, badged computed. /help shows you the verbs privately.
4. Share a piece of the board. Hover a cell on Markets for the ⤴ at its corner, hover the ticker cell for the row, right-click for cell · row · this screen, or press s on a highlighted row. The sheet previews the card, picks the conversation, takes a note, and can quote it as a call. The same ⤴ sits at the end of a row on Trend, Sectors and the other boards, on a Funding heatmap row, on every drawer section header and on each Charts pane (a chart goes as a picture). A shared screen can be live: everyone's copy re-runs your filters on their own snapshot, with newcomers marked and drop-outs struck through.
5. Bind an alert to the conversation. /alert NVDA > 200, /alert NVDA crosses down 180, /alert any rvol > 3 — when it fires, the fire posts here. /alert list, /alert off <id>.
6. Review the record. calls ↗ in the rail, or the record chip beside a caller's name on any call (rank · win rate · closed calls over 90 days), opens the Calls tab — the leaderboard, every call drawn on its candles, and each caller's profile. Deleting a call removes the words, never the stamp or the score.

• Escalation. An unread message reaches your linked Telegram after five minutes — unless you are online, or the thread is muted. An @mention or a message about a ticker on your watch list (the per-person list in the rail, separate from ★) goes immediately and pierces a mute.
• Telegram sync (⇄) in a conversation's header mirrors that one conversation both ways: every message lands on your phone as it happens, and plain text you send the bot posts here under your name. One conversation at a time.
• Browser notifications — enable them in the rail and unread messages arrive with the tab closed.

• Replies quote one line and click back; pins sit in a strip at the top; reactions are a small fixed set; drafts survive a reload; a conversation exports as JSON.
• Attachments: images and plain-text notes, checked by the server itself, not by what the file claims to be. Paste a screenshot straight into the box.
• A call can be armed as a price alert at its stamp — fires when the market crosses back through the level it was called at — or written into your notes book with one click.
• Your usage in the rail shows exactly what the operator can see about how you use the site (which tabs, how long — never what you typed or searched), and has a pause switch.
• The 📣 Calls recap topic posts the week's calls every Monday; join it from the rail if you want it in your list.

1. The ticker — a $TICKER makes the message a call. A message with none, an edit, a command result or a shared card never asks.
2. The side — short, sell, fade, bearish, puts and a few more make it a short; long / buy a long. Options read as a trader would: buy $X puts is short, sell $X puts is long. If your words do not say, the confirmation asks which way? — a call without a side is not a call.
3. The horizon — right after the ticker: $HOOD 30d, 2w, next week, eom, by Oct 15. Unsaid, the confirmation offers 7 days or 30 days. The call closes at that day's close.
4. A target — a price and a deadline after the ticker: $INTC to 32 by Oct 15, with an optional stop: unless 27. It resolves hit, missed or wrong (the stop came first). On a US stock a hit needs a touch in the cash session, or a 5-minute close through the level off-hours; a date deadline ends at that day's 16:00 ET close.
5. The preview names the word it read — "short because of “fade” · horizon from “by Oct 15”". If a target was refused (no deadline, on the wrong side of the mark) it sends as a plain call and the preview says why — never a wrong target.

The weight feeds the leaderboard's rank, its 8-call eligibility and the copy score's sample. Win rate and averages are plain.

• The thesis is sealed with the price. Why (up to 1,000 characters), a catalyst, "wrong if", conviction 1–3 — never editable, and it stays with the call if the message is deleted.
• + update appends a dated note while the call is open (up to 10). Nothing is rewritten.
• +7d extends the horizon a week at a time. It can never be shortened — that is what closing is for.
• close call (in the thread) or close now (on the Calls tab) locks the result at the live mark and asks why are you getting out? — the answer becomes the call's post-mortem. The record tags it ✋ closed by hand.
• write post-mortem — one per call, once it has closed. If your call had a thesis, you get one reminder on your phone within three days of the close.
• ⚑ alert on the stamp arms a price alert at the level you called.

1. Leaderboard — one row per caller, scored on closed calls in the window, direction-adjusted, one unit per call. vs bench is the result minus the S&P 500 proxy (BTC for crypto) over the same hold — "did they beat just being long the market?". Fewer than 8 closed calls (weighted) is provisional: shown, never ranked. Any column re-sorts.
2. Recent calls — every call, newest first. The public record: a call from a channel shows whole; a call from a DM shows the call and its thesis but not the words or who it was said to.
3. By ticker — one market, every caller's calls on it in their own colour, a lane per call. Click a name to isolate it.
4. My copy portfolio — one paper $10,000 account copying everyone you follow, from the day you followed them, with fees and funding in.
5. Insights — the open consensus (two or more callers on one market this week), the record of crowded vs lone calls, calibration (does a conviction-3 call really win more than a 1?) and a head-to-head between any two callers.
6. Weekly recap — one week of the public record: who called what, the best and worst closes, the hand-closes with their reasons, how targets resolved, callers of the week. ◀ ▶ walk the weeks. The same recap posts to the 📣 Calls recap topic every Monday.

• Tiles, the curve against the benchmark, the spread of results, and breakdowns by side, conviction, thesis vs none, horizon, sector and ticker — with a finding where the numbers support one ("conviction is calibrated", "mostly the tape", "banks winners, rides losers").
• The track record, filterable; expand any call for its thesis, updates, post-mortem and result. ✋ marks a call closed by hand.
• The copy panel — what a follower would have made: $10,000 beside the caller's own record, entering 15 minutes late by default (0 / 1h / 1 day selectable), 5 bps a side plus funding.

• Click any call to draw it: the hold band, the entry at its exact minute and price, target and stop, the exit, and lettered pins where the Thesis, each Update and the PM post-mortem were written. 1D / 4H switch candles; vs overlays the benchmark rebased to the entry.
• The timeline beside the chart is a replay: pick an entry and every later candle greys out, so you see what the author saw when they wrote it. ← → step through.

The ring beside each caller (0–100) answers "would copying this person have paid?" — a different question from "are they right?". Every public call of the last 180 days is replayed as a follower would have taken it: entered late, sized 10% of equity, fees and funding paid, total exposure capped. It weighs risk-adjusted return (50%), drawdown (20%), consistency (15%) and followability (15%) — how much of the edge survives an hour's delay. A caller who is right in the first five minutes and flat by the time you read it scores low, and should. Grey means provisional.

• ☆ follow sends a caller's next public call, with its thesis and copy score, to your Telegram and browser push (at most 4 an hour per caller). Never a DM's words.
• You can close your own open call from its row with close now — same dialog as in the thread.
• Calls made in a DM before calls went public stay with the two people in them and are off the public record.

1. Filters › Positions: paste your Hyperliquid wallet's public address and click Link.
2. A ⬡ appears beside every name you hold, in the side's colour; the Position column shows side, notional at the live mark and the move since entry.
3. ⬡ held in Filters narrows the table to your book.
4. In the drawer, the Position panel shows size, entry, leverage, liquidation, margin, funding paid since open, and unrealised P&L and ROE off the live mark.

• Public address only. The app reads the public clearinghouse state and can never sign. It cannot trade for you, and nothing you paste here could let it.
• Your wallet is polled only while somebody is looking at it; the numbers you see move with the live mark between polls.

screenshot — the real app

12

1. 1type a question here
2. 2the answer

1. Ctrl/⌘+K, type a ticker or a tab name.
2. Enter opens the drawer (or the tab). Shift+Enter opens the AI report for that ticker.

1. Press ~ or click ▸_ Ask at the bottom right.
2. Type a ticker for its card, ticker field for one number, or a verb:

• Tab completes verbs, fields and tickers. ↑ ↓ walk your history. help shows the verb card.
• Plain-English aliases work: apr → funding, sqz → squeeze, week → d7, ytd → vs year open, relvol → rvol.
• Anything the local grammar cannot parse goes to the AI leg, which is metered for the group. The ask chip shows what is left today. Local verbs never spend it.

screenshot — the real app

1. Click the 🔔 bell in the status line, open Rules.
2. Ticker (leave blank for any market) · metric · above / below · threshold. Thresholds are in the metric's own unit: 5 means 5%, 50 means 50M.
3. Save. It is evaluated on the server against the same snapshot the board renders, so it fires with no tab open. One crossing is one alert (there is a cooldown), and you can hold up to 60.

screenshot — the real app

1

1. 1the bell panel

1. Bell → Delivery → link. The panel shows a single-use code (ten minutes).
2. Send /start CODE to the deployment's bot. It replies Linked.
3. Pick the alert classes you want on your phone and set quiet hours (a local range like 23-7). test sends a probe.

• On by default: new setups on the actionable board (with your own EV / reward:risk / lateness thresholds and per-name mutes), ledger outcomes (a void taken, a target reached, a horizon resolved), and your rules.
• Opt in: trend changes, 200-day crosses, material SEC filings, earnings on names you have a claim on, macro prints, a flipped AI stance, tape-wide crowding extremes, feed outages. Each shows how often it actually fires, so you can judge the noise before you subscribe.
• Scheduled sends: the morning brief (daily, at your hour), the Landscape commentary a few times a week, and a desk digest — open calls with their levels and close dates, the calls closed this week, the 30-day record per person with copy scores, the open consensus, the newest signals, earnings today and the day's movers. The digest has no default hour: nothing sends until you pick one. Set the hour and the days per send.
• Calls reach you too: a caller you ☆ follow on the Calls tab sends their next public call to your Telegram and browser push; a call of yours that resolves a target posts its result into the conversation it was made in; and a closed call with a thesis nudges you once to write its post-mortem.

• Recent in the bell is the server's own record — it survives a refresh, a closed tab and a redeploy. Clearing it only hides it in this browser.
• Delivery is paced (a message every few seconds, a cap per hour): held, not dropped, and disclosed in the panel.
• /stop to the bot unlinks, always — control of the Telegram account outranks any session.

1. Install it: browser menu → Add to home screen.
2. Accept the mobile preset the first time (ticker, price, 24h, 7d, funding, ΔOI, volume). Every other column is one toggle away in Columns, and once you arrange it yourself the app never overrides you.
3. Keep a phone layout. Which layout is active is per device by design, so save one called phone and leave it.
4. Turn on notifications in the Messages rail for unread messages with the tab closed, and link Telegram in the bell for alerts.

• Your watchlist and saved layouts are already there — they follow the account.
• The Trend table scrolls sideways inside itself rather than widening the page; EMA Touch puts the Touches feed first.
• Nothing is cached on the device on purpose, so you are never looking at yesterday's page.