A dated read on the live market, published most mornings: what the bracket you are holding is worth hour by hour, the odds it survives, and how long you have. Static by design — every figure is as of the moment it was generated, and carries its sample size.
Friday·August 28, 2026today
Eleven more posts, and only the quiet hours left to make them
Generated Fri, Aug 28, 03:45 ET · count 189 on the 7-day market, 21–28 Aug (closes Fri 28 Aug, 12:00 ET)
Where today's market stands
He is on 189 with 8h40m left. Reaching the next bracket at 200 needs eleven more posts, and all of them have to land before noon Eastern.
That is his slowest window. Between 03:00 and noon Eastern he has managed a median of 6 posts since April, across 150 days.
Eleven-plus happens about a quarter of the time. 27.3% on the window measured to the close, 20.0% on the tighter eight-hour version — call it one morning in four.
The market asks 11.5¢ for it. That sits below the bottom of the range the record supports, which is the one number on the board worth a second look.
220 needs thirty-one more. He has done that three times in 150 days. At 0.8¢ the market has it about right.
Why we do not quote the year
His mornings changed in April. Before it, a third to two-fifths of each day's posts landed before noon Eastern. Since, about a fifth — five straight months.
The whole of 2026 says 42.9%. Since April the same window says 27.3%. The year-long figure is five months stale.
It is not a measurement change. The step lands in April, two months before we changed where the counts come from, and nothing moved at that later switch.
Volume did not fall, timing shifted. His daily total has been flat near 29 since May. What moved is when in the day the posts happen.
Zero-morning days are common. 8% of days since April produced nothing at all in this window.
Market 21–28 Aug at 189 with 8h40m to run. Bracket odds from the 03:00–noon ET window, 150 days since 31 March.
Caveat. These are unconditional base rates for the hours this market has left, not a forecast of today. The range on each bracket comes from two defensible definitions of the remaining window — the eight full hours from 04:00, and the nine that run to the noon close — and the truth sits between them because we hold part of an hour that has already started. Counting main-feed replies puts the count at 190 rather than 189, which moves the requirement to ten and the odds by well under a point; the figure shown matches the site. Percentages are Wilson intervals on 150 days since 31 March, and a base rate is not a signal — it says what usually happens, not what will.
Saturday·August 29, 2026
Twelve more posts, four hours, and he went to bed at five
Generated Sat, Aug 29, 08:05 ET · count 53 on the 2-day market, 27–29 Aug (closes Sat 29 Aug, 12:00 ET)
What has to happen
He is on 53 with four hours left. Twelve more posts push him out of 40–64 and into the next bracket up.
Twelve is not a normal morning. Between 08:00 and noon the median since April is 3 posts, and four mornings in five finish under eight.
It happens on 12% of mornings. Eighteen of 151 days since 31 March produced twelve or more in that window. The 95% range is 8–18%.
He last posted at 05:00. That is late for him — he normally finishes around 2am — and it is the specific thing worth conditioning on today.
The market is charging 5.6¢. Below the unconditional rate, in line with the late-finish one. Both readings are live.
Does the late finish change it
On the point estimate, yes. Mornings following a 4–5am finish cleared twelve once in 21, against 12% across all days.
On the evidence, not measurably. One event in 21 days gives a 95% range of 1–23%, which covers both the base rate and the market price.
Every band crosses the price. No group of days we can cut has an interval that excludes 5.6¢. That is the honest state of it.
The wildest mornings are the ones he never slept. Still posting at 6 or 7am gives the highest rate of clearing twelve and the highest rate of posting nothing at all. Today is not one of those.
Twenty-one days is the limit. This is the thinnest cut on the page and the interval is doing real work. Read the range, not the point.
How often twelve or more posts land between 08:00 and noon, by when he last posted, with 95% intervals. Post-April regime, 151 days.
Caveat. These are base rates for the hours this market has left, not a forecast of today. Stop time is resolved to the hour, so a 05:50 finish and a 05:05 one are the same row. The late-finish band holds 21 days and one qualifying event, which is why its interval runs from 1% to 23% — it cannot distinguish the market's price from the unconditional rate, and saying so is the result rather than a hedge. Percentages are Wilson intervals over the 151 days since 31 March; the faint all-time figures use 257 days back to 14 December, which run higher because his mornings were busier before April. A base rate says what usually happens, not what will.
Sunday
No drop published for this day yet.
Monday·August 31, 2026
The pace says yes, the clock says no
Generated Mon, Aug 31, 02:50 ET · count 53 on the 2-day market, 29–31 Aug (closes Mon 31 Aug, 12:00 ET)
Where today stands
He is on 53 and needs twelve. Twelve more posts before noon Eastern break him into the 65–89 bracket.
His own pace says he makes it. He has run 1.37 posts an hour across this market, and twelve in the nine hours left works out at 1.29.
The clock says otherwise. Those particular nine hours run 0.94 an hour on the post-April record and deliver a median of 6, not 12.
Every cut lands near the price. The odds run 17% to 24% depending on how you slice it, against a market asking 16.5¢, and all four intervals contain that price.
His last post was the 23:00 hour. Three silent hours since — an early night by his standards, and nothing yet today.
The things that might matter
Monday is one of his quieter days. 27.3 posts a day since April, second only to Saturday. Across all of 2026 Mondays average 32.1 with a median of 30.5.
Austin is his slowest base. This window runs 6.98 posts there against 10.19 in San Jose — the same small-hours gap we found when we looked at location directly.
An early night is not a quiet morning. Days that are silent through 00–03, like this one, still deliver a median of 6 in this window. The same as any other day.
Midweek is where his volume lives. Wednesday and Thursday both clear 35 a day. Monday and Tuesday sit near 27, and Saturday is the floor at 24.
He did go to Memphis. Landed 16:00 Sunday, left 21:29, back in Austin by 22:41, with 2 posts across the hours we can place him there. Third visit on record.
The chance of adding twelve before noon, cut four ways with 95% intervals, against the market price — plus average posts per day by weekday since April.
Caveat. Base rates for the hours this market has left, not a forecast of today. The four cuts are not independent — today is simultaneously a Monday, an Austin day and a silent overnight, and we have not stacked them because the sample will not carry it. The Austin figure is the firmest of the three (60 days) and the Monday and overnight cuts hold 21 and 24, so their intervals are wide on purpose. Austin runs quieter than San Jose in this window at Welch t = 2.19, but not reliably quieter than everywhere else pooled, and it is one of several comparisons we have run this week — treat it as consistent with the location work rather than fresh evidence. Percentages are Wilson intervals over the 153 days since 31 March.
Tuesday·September 1, 2026
Two posts end it, and he has always found two from here
Generated Tue, Sep 1, 03:45 ET · count 158 on the 7-day market, 25 Aug – 1 Sep (closes Tue 1 Sep, 12:00 ET)
Whether 140–159 holds
Two posts end it. He is on 158, and staying in the bracket means fewer than two posts across the eight hours to noon.
Ignore today and it is 22%. That is the unconditional post-April rate, and against a market asking 17.5¢ it makes the bracket look cheap.
Condition on the silence and it is 5%. Just 1 of the 19 days that were silent from midnight to 4am went on to produce fewer than two by noon.
The exact match is 0 of 7. Days following a 4–6am session that were also silent overnight. The fewest he managed by noon on any of them was two.
A quiet night means a busier morning. Silent-overnight days run 8.11 posts between 04:00 and noon against 6.82 overall. Silence means asleep, and he wakes up posting.
The other two questions
Skipping both sessions: 4.5%. Seven of 154 days since April had nothing between midnight and 4am and nothing at 10 or 11am either.
The two skips are independent. If unrelated you would expect 6.3 such days; there were 7. Skipping the night tells you nothing about the morning.
Skipping both is not a quiet day. Those seven still averaged 21.7 posts. He moves the volume to other hours rather than losing it.
Tuesday owns the late morning. 4.64 posts at 10 and 11am against 3.40 on an average day — his strongest of the week — though 04:00-to-noon overall is ordinary at 7.05.
Two is a low bar. On 78% of mornings since April he has cleared two posts before noon. The bracket is betting on the other 22%.
The chance 140–159 survives, cut five ways with 95% intervals, against the market price — plus the seven days matching today's exact conditions.
Caveat. Base rates for the hours this market has left, not a forecast. The two cuts that describe today are thin — 19 days and 7 days — and both intervals still contain the market's price, so this is a lean rather than a call. They also point opposite ways before they are combined: yesterday's 4–6am session on its own argues the bracket is more likely to survive, and only reverses once today's silence is added, which is exactly the kind of conditioning that can be read into noise. The 0-of-7 figure is the closest analogue we have and the weakest evidence on the page at the same time. The count of 158 is the official settlement figure. Percentages are Wilson intervals over the 154 days since 31 March.
Wednesday
No drop published for this day yet.
Thursday·September 10, 2026
The market is surer about the middle than the record is
Generated Thu, Sep 10, 03:40 ET · count 135 on the 7-day market, 4–11 Sep (closes Fri 11 Sep, 12:00 ET)
Where the market stands
He is on 135 with 32 hours left. Twenty-five more moves him out of 140–159 and into 160–179. Fewer than that and he stays put.
He has been slow all week. 1.00 posts an hour across this market against a normal 1.16, and 21.4 a day over the last five full days against a usual 30.
The middle bracket looks a shade rich. 160–179 is priced at 58¢ against 49% on the full record and 55% on days following a slow week.
140–159 looks a shade cheap. 19.8¢ against 24–25% on those same two cuts. It has already run up hard, and the record says it is not finished.
120–139 is gone. Holding it needs four posts or fewer in 32 hours. The quietest of 162 windows since April was seven, so 0.3¢ is about right.
What does not move it
The launch is noise. Windows containing a SpaceX launch average 38.7 posts against 35.6 without. Welch t = 1.18 — it leans up and does not survive a test.
Neither does San Jose. 38.9 posts there against 35.3 in Austin, t = 0.87. Where he is has never separated cleanly on a window this long.
Thursday is the weakest line here. It rests on 23 day-pairs and is the only cut that disagrees, putting 180–199 at 34.8% against 16–21% elsewhere.
The median window is 35 posts. That lands him on 170, mid-bracket — which is exactly why the market likes 160–179 as much as it does.
But a quarter come in under 25. And that quarter is the entirety of 140–159. The spread is the whole disagreement.
Every 32-hour window since April, coloured by the bracket it would produce, with three cuts of the record set against the market price.
Caveat. Base rates for the hours this market has left, not a forecast. The window is measured as 04:00 through noon the following day, which is the shape of what remains rather than an arbitrary slice. The three cuts are not independent and should not be stacked: 'slow week' holds 44 day-pairs and 'Thursday' only 23, which is why Thursday's 180–199 figure sits so far from the other two — read it as noise rather than as a fourth opinion. The launch and location tests are both null results on 162 windows, meaning too small to detect at this sample size, not proven absent. Percentages are Wilson intervals.
Requested
Questions people asked, answered against the same bar as everything else on this page — including when the answer is that we cannot tell you yet.
Requested·August 29, 2026
“He normally stops posting around 2am ET. On the nights he goes later — 4am, 5am — does he then skip his usual 10 and 11am posts?”
It is not how late he stopped, it is whether he stopped at all. Moving the finish from 2am to 5am barely changes the odds. Still posting at 6 or 7am nearly doubles the chance the late-morning window is empty.
Generated Sat, Aug 29, 09:30 ET
What the record shows
Stopping later barely matters. He skips 10–11am on 31% of days after a 2–3am finish and 26% after a 4–5am one. That difference is inside the noise.
Not stopping matters a lot. On days he is still posting at 6 or 7am he skips the window 53% of the time, against 28% after a 3am finish.
The trend is real but narrow. Across all stop times the trend test gives z = 3.66. Remove the days he never stopped and it falls to z = 0.93 — nothing.
Inside the sleeping range there is no order. As the finish moves from midnight to 5am the skip rate runs 14, 19, 31, 28, 19, 36 percent. That is not a gradient.
A 5am finish is thin evidence. Only 14 days in 257 end that way, and they sit in the flat part of the curve rather than the cliff.
Why it looks stronger than it is
A month cannot see it. The same relationship measured on any single 30-day window ranges from −0.35 to +0.20 — either conclusion is available depending on the month.
Volume is a different question. How many posts he makes overnight is unrelated to the late morning: correlation −0.005 across 257 days.
Two late nights are not alike. Twenty posts ending at 1am and three ending at 5am look identical by volume and opposite by timing. Only the timing carries any signal.
Averages hide the effect. Mean posts at 10–11am barely move with stop time. What moves is how often the window is empty, which is the thing worth measuring.
Validated back to 14 December. Hourly zero-counts begin then. Earlier, a silent hour cannot be told apart from a missing one.
How often the 10–11am window is empty, by the hour of his last post before it. Every day since 14 December with complete hourly coverage (244 with an identifiable stop time).
Caveat. Corrected 29 August. The first version of this answer measured the wrong variable — it tested how many posts he made overnight rather than how late he was still going, and on that basis reported no effect at all. The question asked was about timing, and on timing there is a real effect, though a narrower one than it first appears: it is carried entirely by the days he never stopped. Stop time is resolved to the hour, so a finish at 05:50 and one at 05:05 are the same row here. The skip measure is whether the 10:00 and 11:00 hours together contain zero posts; mean volume shows no relationship either way. The trend test is Cochran-Armitage across stop hours 00:00 to 07:00. Bands below 06:00 hold 14 to 58 days each, so individual bars move easily — the honest reading is the split between stopped and not stopped, not any single hour.
Requested·August 29, 2026·via Discord
“Does anyone have Elon's post count for 13 October 2023, 25 June 2024 and 14 October 2024?”
22, 25 and 54. All three came in below the pace he was keeping at the time — including the 54, which landed in a month averaging 62 a day.
Generated Sat, Aug 29, 13:15 ET
The counts
13 October 2023: 22 posts. A Friday, and 77% of the 28.6 a day he was averaging in the two months around it.
25 June 2024: 25 posts. A Tuesday, 78% of his 32 a day at the time, and the 21st percentile of 2024.
14 October 2024: 54 posts. A Monday and easily the busiest of the three, yet still only 87% of the 62 a day he was running that month.
None of them beat their own moment. Every one fell short of the pace he was keeping in the weeks either side.
Records reach back to January 2018. Most dates can be answered this way, so it is worth asking rather than assuming we cannot.
Reading them properly
The 54 reads two ways. It is the 67th percentile of all 2024, well above that year's average of 46 — and a slow day against the fortnight either side of it.
October 2024 was exceptional. That month ran 2,044 posts against roughly 830 in each of the other two, so the same number means opposite things.
He posts in bursts. On the 54-post day only 13 of 24 hours carried anything, and 22 of those posts arrived inside two hours.
Quiet days are short, not slow. The two lighter days were active in seven and eight hours. When he is on, the rate is similar — there is just less of it.
Everything here is Eastern. The same clock these markets settle on, so the figures line up with what a market would have counted.
Post counts for the three requested dates, with how each compares to the surrounding two months and to its own year, and the hour-by-hour shape of each day.
Caveat. All three months are complete in our records — every day present, single source, nothing interpolated, taken from the tracker these markets settle against. Percentiles are against every day of that calendar year; the surrounding-average figure uses the thirty days either side. One note on the hourly charts: for this period an hour with no record means he posted nothing in it, which is why those hours are drawn as empty. That convention holds for daily totals but makes any per-hour average over this era unreliable, so we have not published one here.
Requested·August 28, 2026·via Discord
“How far does the next bracket up move before his active hours — and what is the lowest price it has traded at for different numbers of posts required?”
It got cheaper on 17 of 21 final mornings, by a median of 8¢. The quiet hours are when a long shot dies — not because anything happens, but because nothing does.
Generated Fri, Aug 28, 05:20 ET
What happens between 03:00 and 10:00
Cheaper on 17 of 21 final mornings. Median move −8¢. That is 81%, with a 95% range of 60–92% on 21 markets.
It holds in both formats. 8 of 9 seven-day markets and 9 of 12 two-day, with near-identical median moves of −8.6¢ and −7.0¢.
Nothing happens except the clock. No news arrives in those hours. A bracket that needs posts he is not making loses value on its own.
The four that rose are the mornings he woke early. Those repriced hard the other way — up 15.5¢, 17.1¢ and 20¢ once the posts started landing.
The largest single collapse was 79.6¢. A two-day market where the next bracket went from 80¢ at 03:00 to under a cent by 10:00.
What the next bracket costs
Six to twelve hours out, the gap sets the price. Needing 1–4 more posts typically costs 71.5¢; 5–9 costs 49.7¢; 10–14 costs 20.8¢; 15–20 costs 14.1¢.
The floor falls faster than the middle. The cheapest ever recorded runs from 46.9¢ at 1–4 needed down to 1.9¢ at 15–20.
Today sits at the bottom of its band. Eleven needed with the market asking 11.5¢. Across nine comparable markets it never traded below 12.5¢ and usually sat near 20.8¢.
Mid-market prices are not comparable. These settle on where the count lands, not whether it passes a mark, so the next bracket up is cheap early because he is expected to overshoot it. Only the endgame compares.
Nine markets is a thin base. Read the distance table as a range, not a rule. It pools both formats to reach usable counts.
Every final morning we hold prices for (21 markets), and the price of the nearest bracket above the count at 6–12 hours before close.
Caveat. Price capture began on 25 July, so the whole record is five weeks and 22 closed markets — this is the smallest base we publish on and it is shown on every figure. The running count is taken through the last completed hour, so no price is compared against posts that had not yet happened when it traded. The distance table pools 7-day and 2-day markets to reach usable counts, and their bracket widths differ (20 and 25), so treat the bands as approximate. Twenty-one final mornings is a small sample and the four exceptions were large ones — a price pattern is a description of what has happened, not a trading rule, and the direction is mostly the clock rather than anything anyone knows.
Requested·August 27, 2026
“When he is in San Jose versus Austin, what is his average posts per day and per hour?”
Close enough to call a tie — 1.34 posts an hour in San Jose against 1.18 in Austin. What location actually changes is the hour he posts, not the number.
Generated Thu, Aug 27, 13:30 ET
How much he posts
San Jose: 1.34 an hour. 32.3 per 24 hours, across 1,839 tracked hours in 27 separate stays, median full day 30.
Austin: 1.18 an hour. 28.3 per 24 hours, across 1,243 hours in 37 stays, median full day 27.
That gap does not survive a test. On full days spent entirely in one city it is 32.5 against 27.4, Welch t = 1.56 — inside the noise.
Every stay counts, landing to takeoff. On 20 August he leaves San Jose at 03:36 Eastern: 24 of that day's 85 posts go to San Jose, 61 to Austin. Nothing is dropped for being a travel day.
The last ten stays run the other way. San Jose's most recent ten made 297 posts in 204 hours, Austin's 354 in 328. Ten stays is far too few to mean anything — it is shown so you can see the raw runs.
Other cities are too thin to publish. Jackson and Hawthorne have a couple of hundred tracked hours between them. They look dramatic and they are not evidence.
When he posts
The clock travels with him. San Jose peaks at 02:00 Eastern, which is 11pm Pacific. Austin's big hour is 11:00 Eastern, which is 10am Central.
The small hours are the real gap. 00:00–04:59 Eastern runs 10.6 posts on San Jose days against 5.6 on Austin days.
It is sharpest at 02:00–03:59. 5.8 against 2.3, Welch t = 4.07, and it survives correction for the four windows tested.
The close window does not differ. 9.1 posts against 7.5 in the 04:00–noon stretch that decides these markets — well inside the noise.
So it tells you when he is awake. It does not usefully tell you what a market lands on. We would not price off it.
Posts per hour by city, the last ten landing-to-takeoff stays at each, and the hourly profile (1,839 San Jose hours, 1,243 Austin hours, 2026 with jet tracking from 20 Jan). The last Austin row is still open — he landed there at 01:52 Eastern this morning and has not left.
Caveat. Location comes from the jet tracker. He is placed at the city of the aircraft that most recently landed and stays there until that same aircraft takes off again, and every hour inside that window is counted where he was. The jets are not the man, so this is an inference rather than a sighting. Flight coverage starts 20 January. The significance test uses only full days lying wholly inside one stay, which leaves 59 for San Jose and 22 for Austin — the Austin figure is below the sample size we would normally publish, so the hour-level numbers carry the weight. Two earlier versions of this answer were wrong and have been replaced: the first attributed whole days by where the jets ended up, crediting Austin with posts made in San Jose earlier the same day; the second listed only days spent entirely in one city, which silently dropped arrival and departure days such as 20 August.
Requested·August 25, 2026·via Discord
“How many times, and what percentage, have the underdogs (0–10¢) won at the end in 2-day and 7-day markets?”
About 6%. Brackets two or more steps above the leader win 9 times in 159 — and that is what a sub-10¢ bracket almost always is.
Generated Tue, Aug 25, 12:10 ET
The answer
Underdogs win about 5.7%. Nine of 159 resolved markets finished two or more brackets above the one holding the count 8 hours before close. The 95% interval is 3.0–10.4%.
7-day markets run hotter. 8.6% against 3.4% for 2-day markets, because a week leaves more room for the count to travel.
The leader holds 51.6%. Just over half of markets finish in the bracket they were already sitting in with 8 hours to go.
One bracket up wins 41.5%. That is the underdog worth watching, and it is why it never trades under 10¢. A cheap quote on that bracket is the mispricing, not the tail.
Falling back is near-impossible. Only 2 of 159 finished below the leading bracket, and those are late data revisions — the count cannot decrease.
Why we did not answer it from prices
Price capture is four weeks old. Only 19 resolved markets carry the quotes this question needs, against 159 with the tweet history behind them.
Counting all sub-10¢ gives 0.7%. Two wins across 282 brackets — a real number, and a misleading one.
262 of those were already dead. Ninety sat behind a count that only rises; 172 needed more than 40 posts in 8 hours. Average price 0.05¢.
Only 20 were ever winnable. Those won 2 of 20, so 10.0%, but on an interval of 2.8–30.1%. Twenty observations cannot price anything.
Both routes agree. 5.7% from 159 markets sits inside the price-based interval. The structural number is the same answer, measured eight times better.
Where markets finish relative to the bracket their count sat in 8 hours before close (159 resolved 2d/7d markets since 14 Dec, with 95% intervals).Every bracket priced under 10¢ at 8 hours to close, split by whether winning was still arithmetically possible (19 resolved 2d/7d markets, 282 brackets).How often the bracket holding the running count fails to hold, by time to close and by position within the bracket (159 resolved 2d/7d markets since 14 Dec).
Caveat. The 5.7% is measured by bracket position, not by price — it answers "how often does a market finish two or more brackets above where it was", which is what a sub-10¢ bracket almost always represents, but it is a proxy rather than the literal quote. A cheap bracket sitting only one step above the leader is a very different bet at 41.5%, so read your own position before applying the number. The 40-post cutoff used to call a bracket unreachable is a judgement call: the largest 8-hour window in 250 days produced 69, and fewer than 6% of days beat 37. Direct price-based measurement needs roughly three times the history we have, so ask again around late October.