How every part of Morning Bytes works.

A morning page can turn into a wall of numbers. One permanent question under the name tells the reader, and the writer, what the page is for.

Every figure below has to help answer it, or it comes off.

The line under the words “Morning Bytes”: What matters before 9:15 am? It sits beside the SAHI mark. In the reader view the stamp “Pre-market edition”, the date, the weekday and the last completed session sit to its right.

The daily message is a different line. It is in the wine panel (module 3).

The reader learns the page’s job in one line. Because it is a question, everything under it reads as an answer. A trader can scan for the answer instead of reading everything.

The writer gets one test for every field: does it answer a question a trader has before 9:15?

1. The analyst writes the line once, in the edition.
2. The page prints it under the title. The title itself is fixed in the page.
3. The checks require the line. An edition without one is blocked instead of printing an empty line.
4. It does not change from day to day.

• It cannot be missing. A missing heading blocks the whole edition.
• “Permanent” is a working rule. Nothing checks that the line reads exactly this way.
• The “9:15” is plain text. On a day when the session opens at another time, the analyst changes the text by hand. Nothing checks it.

A morning page mixes two clocks: the session that finished yesterday, and the moves since. If they blur, a reader can take an overnight reading for a settled fact.

In the reader view, the masthead’s last line names the date and the completed session: “Thursday · after the 23 Sept session”.

The full view adds a strip under the title with up to three items.

• Domestic: the last completed session.
• Cut-off: the moment the edition stopped taking news. A draft calls it “Evening cut-off”.
• Reading order: Conclusion → evidence → conditions.

The reader sees, before any number, that the index figures come from a finished session. In the full view they also see that everything overnight stops at a stated time.

The stamp stops a sample or unreviewed edition being mistaken for a live one.

1. The date, the previous session and the cut-off are entered in the edition.
2. The weekday is worked out from the date.
3. The stamp depends on the edition’s mode: a sample, a real-data draft or a production edition (module 22).
4. The checks require the dates to be real, the previous session to be earlier, the cut-off not to be after publication, and no observation to be stamped after the cut-off.

• The reader view has no strip and no cut-off. The strip is a full-view element.
• The third strip item is hidden on windows 530 pixels wide or less, in the reading layout. Wider windows show it.
• Fact, Observation and Condition are not printed. Whether to print them is not decided.
• A draft may be prepared before the session date, never after it.

Most pages open with a grid of numbers. Here the reader should know the most important thing before reading anything else.

A trader with two minutes can stop at this panel and still leave with the reading.

The wine panel. It has one headline, a short explanation, and under “Before the bell” up to three numbered takeaways. An analyst writes all of it by hand.

Today’s draft reads: “Inside every range. GIFT points to the rising line.” The three takeaways carry their numbers.

The trader decides at a glance whether the day needs a closer look, and which line to read next. Everyone who reads on knows what the evidence below is for.

1. The analyst writes the headline, the explanation and one to three takeaways.
2. The analyst lists which evidence modules the message depends on.
3. The studio checks the length of every line.
4. It shows the panel only while every listed module is confirmed. If one is stale or missing, the panel says “Interpretation withheld” and the rest of the page stays.

• The wording is free text. The studio checks length, not meaning, and does not check that a takeaway states a condition.
• If the analyst removes a module from the dependency list and it then goes missing, a takeaway can still quote its number. Nothing links the words to the list.
• The reader view draws the panel only. The overnight note and “Read together” (module 15) belong to the full view.
• The rules for the headline are proposed, not decided (module 30).
• Whether “GIFT points to the rising line” reads too close to a forecast is an owner decision.

A single overnight number is easy to misread. The strip shows several markets side by side, each compared with its own previous reading. The reader can see what moved and by how much.

A separate rule (module 4.1) decides which markets earn a place.

A row of three to six tiles. Today’s draft has five: GIFT Nifty, Nikkei 225, S&P 500, Brent crude and USD/INR.

Each reader tile shows a label, an optional short note, the value and the change with an arrow. The full view adds what the value is, the time it was observed, what it is compared with, and one line from the analyst. In a draft the full view titles the strip “Evening cues · refresh before the open”.

The reader sees the overnight picture as calm, mixed or one-sided. A GIFT Nifty change is an indication, not a promise of the open. A futures price sits a little above or below the index itself, and that gap has a name, the basis.

1. The analyst enters each cue: value, previous value, unit, scope, what it is compared with, and its source and time.
2. The page works out the percentage change from the two values.
3. A cue marked “price” shows a rise in green and a fall in red. A cue marked “neutral” stays grey.
4. If the contract changed between the two readings, the percentage is withheld.

• A missing cue shows a dash. Nothing is filled in from a related market.
• A cue with a changed contract shows no percentage.
• A value of zero or less cannot be entered today.
• The reader tile prints no time. Only a short note the analyst adds, such as “Close 18 Sep”, tells the reader a reading is older. Module 4.2 proposes printing every time.
• The full view’s closing line depends on the main message. If the message is withheld, that line says to review each observation separately.
• The US 10-year yield, Europe, the dollar index and other Asian markets are owner decisions (module 4.1).

A strip of overseas numbers can grow into a world dashboard. A stated rule keeps it short. Anyone can then ask the same two questions of any cue, shown or missing.

The rule also explains what is absent. “Left out” means tested and not passed, not forgotten.

A two-part test, given by the owner on 24 Sep 2026. A cue earns its place only if:

• (a) it has a known, explainable route to India’s opening reading, and
• (b) it is settled, or trading, before 9:15 am IST.

Nikkei is shown because Tokyo is one of the largest stock markets and trades from about 5:30 am IST, before India opens. China and Hong Kong are not shown. They fail test (a).

The reader can see why each cue is there. The writer has a fixed test for adding or dropping one, so the strip does not drift. A rejected market can be revisited by asking which test it fails.

1. Test (a). Can the route be said in one plain sentence? Example: India imports most of its oil, so Brent matters to the rupee and to oil users.
2. Test (b). Is the market settled, or trading, before 9:15 am IST?
3. Decide. A cue that passes both may be shown. A cue that fails either is left out.
4. State its age. Passing admits a cue. It does not make the reading fresh. A cue must still say when it was read.

• The rule is editorial. The studio does not enforce it. It only requires three to six cues.
• A proposed third test would say: the reading is from the latest session, or states its date. This is proposed, not decided.
• US 10-year yield, Europe, the dollar index, Kospi and ASX are left with the owner. See the table.
• The publish time is an open decision. An edition finished at about 2 am IST cannot contain Tokyo’s session, because Tokyo opens about 5:30 am IST.

Settled means the market has closed for the session and has a final close. The US session is settled after about 1:30 am IST.

Trading means the market is open and its price is still moving at the moment of reading. Tokyo after about 5:30 am IST, and GIFT Nifty through the evening, are trading.

A trading reading is true only as of its time. A settled reading is final for that session. Both are allowed. The reader must be able to tell them apart.

• Every cue is stored with the time it was observed and what it is compared with.
• The reader tile prints only a short note the analyst adds, for example “Close 18 Sep”.
• The full view prints the observation time and the comparison on every cue.
• Printing the time on every reader tile is proposed (module 4.2).

“Owner decision” means it is not settled. The rule alone does not decide those rows.

These are rules for every cue, whatever today’s values are.

• A cue older than the session must state its date.
• Missing stays missing. A gap is a dash, not a guess.
• Nothing is estimated, and nothing is copied from another instrument.

Worked example, from today’s unreviewed draft. The Nikkei tile reads “Close 18 Sep”. That is an old close. Its date is printed so nobody reads it as an overnight move. It must be refreshed before publication. The same draft’s evening readings for the other cues must be refreshed before the open.

The page is updated by hand each morning. A printed time would be wrong the moment a cue is refreshed, and a row of five times makes the strip busy.

Each cue’s observation time is still stored in the edition. The reader tile leaves it out. The full view and the Sources and timing block print it.

The strip stays quiet. A value that is not from the latest session still says so in words on its own tile, as today’s Nikkei does: “Close 18 Sep”.

A cue may carry a short note. The tile prints the note beside the label. Only a cue that needs one has one.

• A missing cue prints a dash and no note.
• The stored time is never deleted, so a later view can bring it back.

A close alone does not show how the day was traded. A chart alone does not say what would change the reading. Each index gets a chart with its interpretation beside it, so the reader can check the claim.

Three cards: Nifty 50, Bank Nifty and Sensex. Each has a header with the name, a range tag, the source and time, and the official close and change. Then comes a chart with two reference lines. On the right are the reading and its conditions (module 6).

The studio can draw candles from numbers. It can also show the owner’s own chart as a picture in the chart slot, as today’s draft does. Whether a picture may stand in for candles in the index cards is not decided.

The trader sees the shape of the recent sessions, the final close, and both references on one picture. They can see how far the close sits from each.

1. The analyst supplies the chart: candle numbers, or the owner’s picture with its source.
2. The header shows the official close and the change from the previous official close, in points and in percent.
3. The range tag compares the close with the two references: inside range, above the upper reference, or below the lower reference.
4. A caption says what the chart is, so a small gap between the header close and the last candle is not a surprise.

• Exactly three indices. A missing index blocks the edition.
• In today’s draft the three charts use different timeframes: 30-minute, daily and 30-minute closes. Each card says which.
• A level read from a picture is marked with “≈” and a tolerance, for example “within about 2 points”.
• A stale index keeps its numbers and shows “Stale chart · review required”. Its wording is withheld.
• Header and chart can differ slightly. The header uses the official close. The chart uses what was supplied.
• Some text inside the picture is small. Read the figure from the card, not the picture.

A previous high is a boundary. Calling it “resistance” claims sellers wait there, which nobody has shown. So the page says reference, records how each was found, and attaches a condition. A line with no condition can never be wrong, so it is not on the page.

The right-hand side of each index card. It holds a one-line view from the analyst. It holds an upper reference (R1) and a lower reference (S1) with prices, a condition for each, and how each level was found. Secondary levels (R2, S2) are optional.

The reader gets two numbers and the exact event that would change the reading: a completed close above or below, on the stated timeframe. Inside the two, no directional exit is confirmed.

1. The analyst enters R1 and S1 with a price and a basis. R2 and S2 are optional.
2. The analyst writes what a close above, inside or below means, in a few words.
3. The page prints the price and the condition for each side.
4. A conditional reading carries its own invalidation line.

• Worked example, from today’s draft. Nifty’s upper reference is 23,470, the flat top of a triangle. The close is 23.2 points under it. The lower reference 23,369 is the rising line. The close is 77.8 points above it.
• The studio requires the lower reference to be below the upper one. It does not check the wording of the conditions.
• Only the basis of R1 and S1 is printed. A basis for R2 and S2 is required but not shown.
• Which close counts as “inside” depends on the source: the range tag uses the chart’s last close, not the official close. Near a reference they can differ.
• The rule that levels are “references” unless there is evidence for support or resistance is decided. It is a writing rule, not something the studio enforces.

The commonest misreading of a level is to treat a brief touch as a break. One bar says once that only a completed close counts, and what would undo it.

A wine bar under the three index cards: “A candle close, not a price touch.” with the invalidation: a completed close back inside the range invalidates that range-exit reading.

The reader learns the rule once instead of three times. The timeframe is printed beside each condition.

The text is part of the page, not the edition. If every index is a range reading it uses the sentence above. If any is conditional it tells the reader to use each index’s own rule.

• This is a full-view element. The current reader view does not draw it.
• The reader view therefore relies on each card’s own condition lines.
• Whether “completed close” should be defined on the page itself is not decided.

On expiry day the whole option chain matters more than usual. One tile gives the ladder and the headline figures, so the trader does not open the chain to find them.

The first evidence tile, printed only when the edition carries an expiring contract. Today’s draft shows Sensex weekly options, flagged “Expires today” (or “N days left”).

Left: a ladder of 13 strikes, six either side of the at-the-money strike. Puts are on the left and calls on the right, in lakh (1 lakh is 100,000 contracts). Right: spot, the at-the-money strike, max pain, put/call by open interest and by volume, and at-the-money implied volatility. Then the total call and put open interest. Then four lines: highest call open interest, highest put open interest, most calls added and most puts added.

When both at-the-money prices exist, it also states the range the call and put together price in, “not a forecast”.

The trader sees where positions are stacked and what the market charges for today’s move, in one place. The tile says that concentration is not a guaranteed barrier and that the range is not a forecast.

1. The analyst saves the whole chain of the expiring contract with spot, the at-the-money strike and the totals.
2. The page picks the strikes around the money and works out the highest and most-added strikes.
3. Max pain, the put/call ratios and implied volatility are read from the saved chain. A missing figure is left out.
4. The range is shown only when both at-the-money prices exist.

• The tile does not appear on an edition with no expiring contract.
• Which contract counts when several expire on one day is not decided.
• A day-count flag reads “Expires today” at zero days left.
• Open interest is as of the previous close. Additions are since the previous close.
• If the source is marked missing, the tile says so and shows no figures.

An index can rise because a few heavy stocks rose while most fell. A count of advances and declines shows whether the move was shared. It is the quickest check on how much weight an index move deserves.

A tile titled “Market participation”, with the group counted in its label: “50-stock basket” in today’s draft. It shows the share that advanced, a three-colour bar, and the three counts: up, flat and down.

Today: 74% advanced, 37 up and 13 down.

One glance answers “did most stocks go up?”. The bar and the counts show the split. A 74% figure is never read without knowing how many fell.

1. The analyst enters the group name, the counts of advancing, declining and unchanged stocks, and the total.
2. The studio checks that the three counts add up to the total. If not, the edition is blocked.
3. The page shows advances divided by the total as a whole percentage, and draws the bar in proportion.

• It counts one named group. Today it is a 50-stock basket, not the whole exchange. The full-view card says so in its note.
• Missing: the tile says “Data unavailable”. Stale: the numbers stay with a “stale” label.
• If participation is a dependency of the main message and it is missing or stale, the message is withheld.
• The card is titled “Market participation”. “Breadth” is the everyday word.
• Which group a production edition counts is not decided.

Price alone does not say who stood behind it. Reported cash flows show whether foreign and domestic institutions were net buyers or sellers, and whether they pulled the same way.

A tile with two panels, FII/FPI and DII, in rupees crore net. Each shows the last session as a large figure and five sessions as small bars. Under the bars: the five-session net.

Today: FII/FPI +1,617 and DII +2,342, with five-session nets of −5,378 and +13,896.

The reader sees one group buying while the other sold, or not, and by how much, on one scale. The five-session total shows whether the last day was a one-off.

1. The analyst enters gross buy and gross sell for each group, and up to five sessions of net figures.
2. The page works out each net as buy minus sell.
3. Bars share one scale, centred on zero, so their lengths compare.
4. The checks confirm the latest history entry equals that session’s buy minus sell.

• Only net figures are printed. Gross buy and sell are checked, not shown.
• Figures are provisional and rounded by the publisher.
• Money moved is not a forecast of direction.
• Missing: “Data unavailable”. Stale: figures stay with a “stale” label and the note tells the reader not to present them as the latest session.
• The checks do not confirm the sessions in the bars are consecutive.

A “mixed” market can hide a clear split between sectors. A short list of selected sectors shows the shape of the day without listing every sector.

A tile of up to six sectors, each with its change and a bar on a common scale. Today: Metal +2.40%, FMCG +1.32%, Pharma +0.90%, IT −0.87%. The reader tile prints two decimals. The full-view card prints one.

The reader sees leaders and laggards at once, so a small move is not made to look large. The trader knows which sector names to open first.

1. The analyst chooses one to six sectors and enters each change.
2. Bars run from a zero line: right for a rise, left for a fall.
3. The largest move sets the scale.
4. Rows keep the analyst’s order. The page does not sort.

• These are selected sectors, not a full ranking. Do not call it a sector ranking.
• How the sectors are chosen each day is not decided.
• Missing: “Data unavailable”. Stale: bars stay with a “stale” label.

Cash flows show what institutions did in shares. Derivatives turnover shows another part of what they did. The tile says plainly that it is not a direction.

A tile named “FII derivatives · net turnover, ₹ crore”. Two rows: index futures and index options, each buy value minus sell value. Today: index futures +679.0 and index options −3,906.2. Its note reads: buy minus sell turnover; not open interest, not positioning.

The trader gets the second half of the flow picture without a false positioning read.

1. The analyst enters buy and sell values for index futures and index options, in rupees crore.
2. The page prints the difference for each, to one decimal.
3. The unit must be rupees crore, not contracts.

• The block is optional. Without it, no tile is printed.
• In the full view the same figures fill the futures card’s slot, but only when the futures card is missing.
• FII index-futures long and short positions are on the pre-market checklist as “Not yet” (module 31).

Open interest counts contracts that are still open. Rising open interest with rising price means something different from rising open interest with falling price. Neither can be read alone, so the card shows both.

A full-view card with two changes for one futures contract: the price change and the open-interest change, in percent, with the contract, its expiry and the two dates compared.

The reader can judge whether a move had new positioning behind it. The card reminds them that open interest alone is not a direction.

1. The analyst enters the contract, expiry, and today’s and the previous price and open interest.
2. If both readings are the same contract, the page prints each change as a percentage.
3. The price change is coloured. The open-interest change is always grey.
4. If the contract changed, both changes are withheld.

• In today’s draft this block is marked missing. It is withheld until it is verified, and the reader view has no futures tile.
• A previous open interest of zero gives “N/A”.
• The card prints only the two percentages, not the raw price or open interest.

Many traders watch the strikes where the most option contracts are open. One line gives that without turning the page into an option chain. It reports the highest total, not the biggest change, so the two are never confused.

A line under the evidence cards. It names the underlying and expiry, and says whether the count covers the whole expiry. It gives the strike with the highest open interest on the Call side (CE) and on the Put side (PE). It ends with the scope, the time and the reminder “Concentration is not a guaranteed price barrier.”

The reader gets two strikes to watch and an honest statement of how complete the data behind them is.

1. The analyst supplies the total open Call and Put contracts for each strike, and says whether the list covers the whole expiry.
2. The page finds the highest Call total and the highest Put total. Ties are all listed.
3. The label says “highest total OI” for a complete list, or “highest OI within supplied subset” otherwise.

• This is a full-view line. The current reader view does not draw it. It uses the expiry-day tile instead (module 8).
• It prints strikes only, not the open-interest totals.
• It uses CE and PE. Whether to change to Call and Put is not decided.
• Put/call ratio and max pain are not part of this line.

Evidence rarely points one way. A line that must state both what supports the reading and what challenges it stops the page from quietly picking a side.

One line under the options line, labelled “Read together”. It has two analyst sentences: what supports the reading, then what challenges it. There is no score and no meter.

The trader sees the strongest reason for the reading and the strongest reason against it, side by side, and can check each against the cards.

1. The analyst writes “supports” and “challenges”.
2. The page shows them one after the other, only while the main message is available.

• Full view only. The current reader view does not draw it.
• If a dependency is stale or missing, the line reads “Combined interpretation withheld until the referenced evidence is refreshed.”
• The studio does not check that the sentences agree with the cards.

The owner already draws patterns on Sahi’s own charts. A drawn pattern is often clearer than a list of candle numbers, if it says where it is confirmed or cancelled.

Built in studio 2.2.0. A section, titled “Stocks to watch” in today’s draft, of one to six cards. Each card has the name, exchange, timeframe and span of the pattern, the last price and its date, and the owner’s chart as a picture with its source.

Under the chart the reader view prints one line: the breakout condition. Example: “A daily close above 9,070.50 confirms the exit from the range. A close back under 9,050 afterwards voids it.” The fuller reads stay in the edition data.

The full view also prints each level, how far the last price sits from resistance and support, and three conditions: Confirms, Voids it and Downside.

The reader sees the pattern where it formed, and is told in numbers where it is confirmed or cancelled. The trader knows which stocks to open first and the exact close that would confirm or cancel each pattern.

1. The owner sends a chart picture and names the stock.
2. The analyst confirms the instrument from the candles on the picture, matched to saved bars, not from the label it arrived under.
3. The analyst writes each level with how it was found, and the conditions.
4. The picture is stored inside the edition file, so the edition stays self-contained.

• A level read from the picture is marked “≈” and states its tolerance. A level matched to saved data is exact.
• The identity check matters. The first chart received carried the wrong name.
• When the drawn line and the data highs differ by a few points, the condition is set at a level printed on the chart.
• The rule “a confirmation is a completed daily close beyond an unambiguous level” is proposed, as are the rules that every level states how it was found.
• Not decided: whether pictures may stand in for candles in the index cards. Not decided: which line defines a range top when the drawn line and the data disagree. Not decided: how many cards a day, and whether the checks may verify a picture.
• On a phone the chart can be swiped sideways to read all of it.

A list that is always filled teaches readers to ignore it. So the page allows zero, one or two setups. Each carries the level that confirms it, the level that voids it, the volume context and the event risk.

Full-view cards beside the risk radar. Each shows the name, the last close and a status. It has a small daily chart with volume bars and the trigger line. It has one line naming the setup and one saying why. Two boxes give “Daily close above” and “Invalidate: daily close below”. A volume line and an event check finish the card.

The trader sees what would confirm each name, what would void it, and whether volume backed the last move. When there is nothing, the page says so.

1. The analyst enters each stock: daily candles with volume, close, trigger, invalidation, status and three short lines.
2. The page divides the latest day’s volume by the average of the earlier days and prints it, for example 1.80×.
3. With no setups it prints “No fresh setup selected. There is no requirement to fill the watchlist.”

• Today’s draft has none. It uses the chart cards of module 16 instead. When chart cards exist and there are no setups, this block is left out.
• The template expresses upside setups only.
• How stocks are chosen is not decided. No definition of “today’s criteria” exists yet.
• A stock cannot be missing, and in production it cannot be stale.

A reading can be right and still be broken by a scheduled announcement. The page shows the volatility index as context, lists the events a trader should know about, and is honest when nobody has checked the calendar.

India VIX is a published measure of how much movement the options market expects. In the reader view it is a tile: 10.37, −5.73%, with the note that it is not a direction signal or protection against an opening gap.

Events. The reader view prints an events panel only when a verified event exists. The full view always prints the risk radar: VIX, the events or a calendar status box, a “Before the open” checklist on drafts, and “Read conditions, not certainty.”

The trader gets a calm-or-stretched reading and the exact times of scheduled events. In the full view an unchecked calendar reads “Calendar not verified”, never “no events”.

1. The analyst enters India VIX (value, previous value, note) and the calendar: a status, a note and up to two events, each with name, scope, reason, source and time.
2. The page prints the VIX value and its change in grey, because a rise is not good or bad by itself.
3. Each event prints its time in IST, name, scope and reason.
4. With no events it prints “Calendar reviewed” (verified) or “Calendar not verified” (unverified).

• In today’s draft the calendar is not verified and has no events. The reader view therefore shows no events panel and no sign that the calendar was unchecked. Whether the reader page must carry a “calendar not verified” line is an owner decision.
• A “quiet day” headline is only allowed when the calendar is verified empty.
• Where the calendar comes from, how it is verified, and what counts as a high-priority event are not decided.
• If VIX is marked missing, the full view says so. If it is left out entirely, nothing is drawn.
• Only two events fit.

A page that shows conclusions must show where its evidence came from and when. The footer says what the page is and is not, and who reviewed it.

Reader view: the disclosure, the SAHI web address with the tagline, and “Pre-market edition”. Today’s disclosure says the page is prepared from public market information and Sahi’s own market data, for information only, not advice, and that levels marked ≈ are read from charts and are approximate.

Full view: also “Sources & timing” (every source by id), the line “All times IST”, the review line, the edition id and version, and any correction.

Every source tag in a card resolves here. The review line tells the reader whether a person approved the edition. The disclosure states what the page does not claim.

1. The analyst lists the sources and writes the disclosure.
2. Each card prints its source id. The full-view footer expands the ids into names.
3. The review line reads “Draft · analyst review pending” until the edition is approved. Then it names the reviewer.
4. A correction note, if any, adds a “Correction:” line.

• In today’s draft the sources are public web pages and Sahi’s own saved readings. Data rights, brand and disclosure approvals are all still unconfirmed.
• The reader view does not print the sources or the review line.
• The approved production disclosure and the real source list are not in the product yet.
• Without a source list or a disclosure, the edition is blocked.

A page that quietly shows yesterday’s number as today’s, or fills a gap with a guess, cannot be trusted. Every figure carries its source and time, and a status the reader can see.

A small record attached to every observation: which source it came from (a short id), when it was observed, what it is compared with and when, and a status: confirmed, stale or missing.

Each card shows its source id and time, so the reader can see how fresh it is. A stale or missing status is printed beside the time. The full view’s “Sources & timing” window lists every observation.

1. The analyst gives each observation a source, an observed time, an optional comparison time and a status.
2. The page prints the source id in brackets and the time. A status other than confirmed prints as a small tag.
3. A missing module shows “Data unavailable” and no conclusion is drawn from it.
4. The main message reads the statuses: it needs every module it depends on to be confirmed.

• Stale data is still drawn, with its label. What the page withholds is the conclusion that rests on it.
• Only the index charts and cash flows are held to the previous session’s date. Marking other old data stale is up to the analyst.
• A missing module that still has an old time prints that time beside “missing”. The time should be left out.
• How old is too old is not decided. The studio accepts what the analyst declares.
• A missing India VIX left out of an edition draws nothing at all.

A morning page has to be right, and a tired reviewer’s memory is not enough. The most damaging mistakes are a sum that does not add up, a stale chart shown as current, and two different contracts compared. The checks run before the page can be drawn or exported.

About a hundred checks, run when an edition is loaded or edited. An error blocks the edition. A warning is shown and needs a person’s judgement. A last check looks for text running out of its box.

A bad edition never looks finished. It shows a list of what is wrong, in plain words, and the export buttons stay off. The analyst fixes the data, not the picture.

1. The studio runs the checks on load and on every edit.
2. With errors, it shows “Edition blocked — data needs correction” and the first errors. It states that no previous report is being shown as the new edition.
3. Without errors, the page is drawn and the layout check runs. A layout problem blocks an export.
4. “Validate edition” opens a summary of warnings and review items.

• The checks cover: text length, arithmetic, comparison times, chart and time order, reference order, and stock levels.
• They also cover: a main message that depends on modules that are not confirmed, and the stricter production rules.
• They do not check whether a source is authentic, whether the contract or expiry is right, whether the calendar is right, data rights, brand approval, or whether the analyst’s words are wise.
• The rules for the daily line’s wording are proposed and not run by the studio (module 30).
• The checks test structure and arithmetic. They do not judge a reading.
• Real-data drafts always carry one warning: “Evening snapshot, not morning-final.”

A design example must never be mistaken for a market report, and a real edition must not go out unreviewed. The kind of edition is written into it, shown on the page and enforced by the checks.

Three kinds. A sample is invented and says so in four places. A draft is real data that is not reviewed (module 23). A production edition is real and approved by a named reviewer.

The reader can tell from the page whether it is an example or a reviewed edition. A production edition cannot be drawn until a reviewer has approved it and the sources, brand and calendar are real.

1. The kind is stored in the edition.
2. It changes the stamp, the footer words, the studio’s status text, how strict the checks are and the tags in the archive.
3. A production edition passes the same checks and then stricter ones.

• Production needs: an approved status with a named reviewer whose review time falls between the cut-off and publication; data rights, brand and disclosure confirmed; no made-up sources; no demo identifiers; a verified calendar.
• Only production and approved editions count as published. Everything else is a draft.
• The approved production disclosure, the real sources and who reviews are not decided.
• The studio does not lock an edition. “Frozen” is a rule the team follows.

A sample is invented. A production edition is reviewed. A draft sits between: the numbers are real, the review is not done. It needs its own honest label so it is never read as either.

An edition whose mode is “draft”. The studio bar reads “REAL-DATA DRAFT”. In the full view the stamp reads “REAL-DATA DRAFT · REFRESH BEFORE OPEN” and the footer says “Real observations · not a streaming feed”. Today’s edition is one.

Real numbers can be laid out and checked in advance without being passed off as a released report. The analyst and the owner know what still needs a morning refresh.

1. The analyst sets the edition’s mode to draft.
2. The checks add one standing warning and refuse made-up sources and demo identifiers.
3. A draft may be prepared before the session date, never after it.
4. Review status stays draft, with no reviewer, and the rights, brand and disclosure approvals stay unconfirmed until someone signs them.

• A draft is never published. The site build does not copy it.
• It must be refreshed before the open.
• A draft is never called published.
• Who reviews, and what a review checks, is not decided.
• A missing block stays missing. Today the futures block is withheld.

The reader page is the one that is shared. Workflow notes such as “calendar not verified” or a source list belong to the people who prepare it, not to the reader.

The reader view shows the stamp “Pre-market edition” and a short disclosure. Status, sources, times and the review block stay in the edition and in the full view.

A draft is kept apart by the studio bar, which says so, by the Output page, and by the site build, which never ships a draft edition. Sharing a draft is a deliberate act, as with today’s card.

• Production still needs an approved status, a named reviewer and confirmed rights before the studio will call an edition reviewed.
• Who reviews is not decided.

Most people meet a morning page on a phone. A desktop sheet shrunk to fit is unreadable. So the same page, from the same data, is laid out again for reading.

A second layout of the same page. It is used when the window is under 1,024 pixels wide. A “Page view” menu in the studio lets you choose Page view, 100% size or Reading view.

On a 390-pixel phone each index card is one column, with the chart on top and the two conditions side by side under it. Tiles and cards stack. The type is larger. The page does not scroll sideways.

1. When the studio opens, a window under 1,024 pixels gets the reading layout. Otherwise it gets the page view: a 1,200-pixel sheet, scaled down if the window is narrower.
2. In the reading layout the reader-view tiles form one column.
3. In the full view the evidence cards go two across from 531 pixels and one across at 530 pixels and below.
4. Under 620 pixels the chart is drawn taller and its text larger.
5. On chart-picture cards, the picture can be swiped sideways.

• The layout is chosen once, when the page opens. Resizing does not switch it.
• A desktop window dragged narrow stays in the page view, scaled down.
• The studio’s own controls wrap into rows on a phone and are left out of exports.
• The third item of the full view’s time strip is hidden at 530 pixels and below.

If a page were put together from many places, nobody could say what a figure was based on. One file per morning means one thing to write, check, keep and reproduce. The page is only ever a drawing of it.

A single data file that holds every fact and every sentence of one morning, including its own chart data and the owner’s chart pictures. The studio draws the whole page from it.

The analyst has one place to enter figures and a list of what is required. The reviewer can read the file and see exactly what the page will say.

1. The analyst writes the edition and loads it into the studio.
2. The studio checks it and draws the page if it passes.
3. Drawing never rewrites the data. Numbers are rounded only as they are printed.
4. The published file is kept, and the archive lists it.

• Some fields are required but never drawn.
• Nothing locks an edition in the studio. Frozen is a working rule.
• The archive lists an edition if it has a session date. It does not re-run the checks.
• How versions and corrections are numbered is not decided, and neither is the file name of a production edition.
• Field names and limits are in the engineering notes.

The page is often shared as an image. An image made by hand can drift from the reviewed page. Exporting from the page itself, and only after the checks pass, means the picture is the reviewed edition.

The studio’s “Download PNG” button saves the page as one picture, named after the edition id, without the studio controls. “PNG · smaller file (2×)” is in the More menu. Print prints one page sized to the sheet.

One click gives a sharp picture of the whole one-pager. An invalid or overflowing edition cannot be exported.

1. The studio draws the sheet at 1,200 pixels wide with its controls hidden.
2. It refuses if the checks or the layout fail.
3. It saves the file at three times the size, or two times from the smaller-file item.

• If a browser cannot draw the picture, the message says so and points to the engineering route.
• The two routes were not compared pixel by pixel.
• Which route is the published one is not decided.
• A picture of a draft still says “Pre-market edition” (module 24).

A daily publication needs a record. The archive keeps one entry per morning and shows two separate facts: is the data live, and has a reviewer approved it.

The Output page. It counts editions, published and drafts, shows the latest edition large with its headline, status, dates, reviewer and id, and lists earlier ones as cards. Each has links to open it in the studio and to download its data file.

The reader sees at once whether an edition is sample or real, draft or approved, and can open or download any of them.

1. Editions are saved as data files in a folder.
2. A step run by hand rebuilds the list on the Output page.
3. The page sorts by session date and draws it. Nothing is fetched, so it opens straight from disk.

• A new edition does not appear until that step is run.
• The list does not check that an edition passes the checks.
• The public site lists the sample edition. It does not list a draft, and it does not copy the draft’s data file.
• “Open in the studio” needs the folder to be served. Opened from disk, the studio asks you to use Load JSON.
• Where the archive is shown publicly, and which editions appear, is not decided.

The page is drawn from data, so the tools for working with the data live beside the page. They are never part of an exported picture.

The bar at the top of the studio: a status, a Page view menu, and the buttons Edit edition, Load JSON and Download PNG, then a More menu with the smaller PNG, Save HTML, Save JSON, Validate edition, Sources & timing and Print one page.

One place to open an edition, change it, see whether it passes, see where every figure came from, and get a file out.

1. Load JSON opens a file, checks it and draws it, or shows the blocked page.
2. Edit edition shows the data in a box. “Validate & apply” accepts the change only if it passes.
3. Sources & timing lists every source and every observation with its times.
4. Save HTML saves a self-contained copy. Save JSON saves the data.

• Download PNG, Save HTML and Print are off while the edition is invalid.
• Save JSON is not disabled for an invalid edition.
• The status reads “REAL-DATA DRAFT”, “SAMPLE · NOT LIVE”, “REVIEWED EDITION”, or the number of data errors.
• The Edit box does not lock the edition’s mode or review state. It only re-runs the checks.

A hand-written headline can still make two mistakes on a page that promises never to over-claim. It can speak about a state the numbers do not show, such as “inside the range” over a chart that closed outside it. Or it can assert something it cannot support: a direction, a prediction, a number found nowhere in the edition.

Proposed, not decided. The analyst declares which of six kinds of day the headline speaks about. A rule-check would refuse a kind the figures contradict, and would check the wording against language rules. The analyst still writes every word. The studio does not run any of this today.

Today’s edition declares its kind of day as “inside range”. The studio does not check that.

The analyst is stopped from writing “quiet day” when an event is due, and from writing a number that is not in the edition. The reader gets a headline that is always about a state the data supports.

1. The analyst declares a kind of day.
2. The check works out which kinds the figures allow, strongest first.
3. A declared kind that does not apply is an error. A stronger kind that applies is a warning.
4. The check tests each line for banned words, advice, predictions, probabilities, times and numbers.

• Not decided: whether the check joins the studio, and whether the range should always be judged by the official close or the chart’s last close.
• A rule-check that rejects “will” even in “will not” shows how strict the number and word rules are.

Nifty’s official close, moved by GIFT Nifty’s own percentage change. It uses GIFT’s own move, not GIFT minus spot, because GIFT is a futures price with a gap to spot.

Sample figures: GIFT 24,825 to 24,850 is +0.100705%, so a close of 24,820 becomes about 24,845. It is an indication, never a forecast. It is absent when the GIFT reading is missing or the contract changed.

• No direction words, no advice, no prediction verbs, no probabilities.
• Every number must be in the edition or be arithmetic on it. For example 24,900 − 24,820 = 80.
• Every time must be 9:15 or a verified event’s time.
• Each line has a hard length limit and a softer “reads best under” length.

A page earns its place by answering the questions a trader has before 9:15. Laying those questions out in the order a trader asks them shows what the page covers and what it leaves out.

Proposed: a list of checks, each with the question it answers and whether the page carries it. It exists as a working list, not on the page. Whether it becomes visible cards is not decided.

The team can see the gaps. A rule decides what joins the page: a row joins only if it changes the reading.

The analyst works down the list. Every figure that is carried is entered by hand with its source and its time.

• On the page today: overnight cues, the three index charts, cash flows, FII derivatives turnover, expiry-day open interest, India VIX, participation, sectors, chart patterns to watch, and the event calendar.
• Not yet, all proposals: the US 10-year yield, FII index-futures long and short, and index heavyweights (was the index carried by two or three stocks?).
• Every “Not yet” row is undecided.