MyStatus Stats

Make your decisions now with the facts...and there’s a lot of facts.

How to read these numbers

  • Every figure names its window. If it is not counted from the first commit on Jul 26, 2026, it carries a Since date. No date means all of it.
  • Everything floors. Hours, sessions and averages round down, and where a rounding could go either way it goes the way that does not flatter us.
  • Stated is not measured. A figure nothing heremeasures is stamped Stated, not measured where it stands. It is somebody’s word, and it should cost you more to believe than the ones with a record behind them.
  • Hand-kept records say they are hand-kept, and name the day they were last checked.
  • The signs read a record, not a mood. A sign is lit because a label and a timestamp say so on a public task you can open — not because anyone is typing. Nothing on this page is scraped out of prose.

The AI Stats — Token breakdowns

Before any work starts, an assistant is handed a fixed block of standing context: its own instructions, this project’s conventions, the description of every installed skill, the signature of every available tool. None of it is the work. All of it is re-read on the next turn, and the one after that.

106,088
Tokens before a word is written

Handed over on every single turn, whatever the turn is about.

52%
Of it we don’t control

The assistant’s own system prompt. The largest line item is the one nobody in this project can edit.

6,993,297–14,743,215
Per finished task

Across 3 finished tasks, each multiplied by its own floor — the figure drifts as the toolchain does.

  • The assistant’s own system promptby subtraction55,84452%

    Anthropic’s standing instructions, not readable from inside. This is what is left after everything else is subtracted, so it also carries the error in every estimate above it — read it as a ceiling on the part nobody here can edit, rather than an exact figure. It is still comfortably the largest line, and its share went DOWN this week.

  • Project instructionscounted18,42417%

    This repo’s CLAUDE.md. Conventions, invariants, and every gotcha learned the hard way — the file that lets a session pick up where the last one stopped. It grew by more than six thousand tokens in a week, which is the single largest thing we did to our own bill.

  • Tool definitionsestimated14,50013%

    The full call signature of every tool loaded up front. One of them is 6,800 tokens on its own — as much as this entire line was estimated at a week ago. Tools held in reserve cost only their name until they are actually needed.

  • Skill descriptionsestimated12,10611%

    52 installed skills. Each description is presented twice, in two different formats, so every skill costs about double its face value. The only line here that went DOWN this week, by dropping three skills.

  • Connector instructionscounted2,5712%

    How to use each connected service. Two of the connected servers here are duplicates of each other, and the duplicate’s instructions are paid for twice.

  • Memory indexcounted1,7361%

    One line per stored fact about this project, carried into every session. 39 of them now.

  • Reserved tool namescounted9070%

    87 tools listed by name only, loaded on demand. The cheapest thing in the list, and the reason the tool count is not the problem people assume — the line above is.

The total is measured; the breakdown is not, evenly. The 106,088 is exact, read out of a session's own transcript. The parts above are good to about ten or fifteen percent, so trust their order rather than their digits.

A dated snapshot, not a live counter: the transcripts it reads sit on one laptop and never enter the repository, so one person refreshes this by hand. Last on Aug-30-26.

The re-count went the other way from the easy fix. Every part above was measured again on Aug-30-26. Simply raising the total would have pushed the share we do not control from 56% to 65% — publishing that the untouchable part is what grew. Counted properly it fell to 52%. The week's two largest additions were this project's own conventions file and the tool signatures loaded up front, and both of those are ours.

And it keeps going up.

Every coding branch and every thinking session opens with a public entry carrying its own figure. Lined up, those entries are a price history for this project’s own rulebook — and every edit to it becomes a decision with a number attached.

110,00080,000
Aug-24-26Aug-30-26
Coding sessionsBusiness sessions
+25%
Growth in a week

From 84,647 to 106,088 tokens on every turn. Nothing was deleted to make room.

9,515
Added in one evening

The biggest single step, and 44% of the whole week's growth. Two sessions on Aug-28-26, hours apart.

2,144,100
What the rise costs a 100-turn task

The increase multiplied out. Not a task that got dearer — the same task never runs twice — just what the week added to a job of that length.

Every point is a published entry, not a re-measurement. Each figure was read out of the Work Log or the business log at the moment a session opened, and the row it came from is still there to check. A session quotes the same figure on every branch it opens, so a run of identical entries counts once: 53 readings out of the entries published so far.

The axis starts at 80,000, not at zero — the whole week fits inside a fifth of its own size, and a zero baseline would draw it as one flat line. The two kinds of session load slightly different toolchains: the widest gap between a business reading and the coding reading nearest it in time is 1,195 tokens, in both directions across the week. Read the climb; ignore the wobble.

Measured task by task, not estimated. Every task closes with its own turn count and scaffolding figure on the MyStatus Work Log, so the figure above is checkable against its source rather than asserted here. There are 3 so far, which is why it reads as a range.

  • Token count in prompts14,743,215

    177 turns at 83,295 each. Closed before the receipts named a model. The entry it came from

  • Message editing and token usage11,935,227

    141 turns at 84,647 each. Closed before the receipts named a model. The entry it came from

  • Model name in the logbook’s token receiptsclaude-opus-56,993,297

    81 turns at 86,337 each. The entry it came from

The model is a receipt, not an explanation. Naming it removes one thing a reader would otherwise have to guess at when two rows here disagree — different models use very different amounts — and that is the whole of what it claims. It does not explain the variance: these tasks also differ in how big they truly were and in how much of each one went on debugging, and no figure on a row separates those. It records what ran, not a choice anyone made; a session’s model is set outside the conversation. Rows that closed before the entries carried a model do not have one here either.

What a turn costs as a session gets older

Every turn re-reads the conversation so far, so the same work costs more the later it lands. Measured across 144 sessions and 14,640 turns of this project's own transcripts, each turn priced against its own session's first 20 — so a session is only ever compared with itself.

  • 1-201.00x
  • 21-401.13x
  • 41-601.29x
  • 61-801.46x
  • 81-1001.63x
  • 101-1401.85x
  • 141-1802.10x
  • 181+2.86x
  • scale ends at 3.00x

The threshold is turn 141. That is the first stretch where a turn costs 2.1x what an early turn in the same conversation cost. It is published with its arithmetic rather than as a rule, because a fresh session is not free either: at that ratio, each further turn on the old seat wastes about 1.1 of a fresh one, so starting over pays for itself after roughly 10 more turns of work when the handover costs ten.

The long sessions did not start out expensive. Only a long session has a hundredth turn, so the right of this chart is a different set of conversations from the left, and a rise there could be nothing but that. The 16 sessions that ran furthest opened at a median 18,868 tokens a turn, against 18,580 for the whole set. They started at the ordinary rate and got expensive on the way.

Hours on the record

Two things a clock can answer: how long the work was open, and whose time it was. Every session is a sequence of timestamped records, so the wall clock splits in two — time the assistant was producing, and time it was stopped, waiting on someone to read and decide.

243.5
Hours worked

Since Jul-27-26

That reaches back to within a day of the first commit. Each minute counted once, however many sessions were live in it — 56.8 hours had two.

The model behind this number keeps getting refined, and the refinements are published too. This figure comes from session transcripts, which sit on one laptop and never enter the repository. The public work log — which drives the breakdown in the next section — reads about a third less over the same days, because it only counts work that announced itself with a posted opening and closing. Both stay on the page. Neither is used to correct the other.

Who the clock belonged to

67%
Of session time is the human

Since Jul-27-26

201.3 hours reading and deciding, against 99.0 of the assistant working.

1,641
Stretches where nothing moved

Across 143 sessions. Each one is the work stopped, waiting on a person.

406.1
Thrown away as absence

Any gap longer than 45 minutes is a closed laptop, not deliberation. More was discarded than kept.

Two questions, two denominators. The share is of SESSION time: a minute with one session thinking and another coding is honestly one of each, so the parts sum to 300.3 hours. The headline is the CLOCK, where that same minute is one minute — 243.5. Neither is derived from the other's denominator, and the 56.8 hours between them is time two sessions were live at once.

This is a floor, and it is a floor three times over. It counts only ATTENDED review — the stretches where the work was demonstrably stopped, either just before someone typed or inside a question that suspends until it is answered. Reading along while the assistant works does not count. Thinking away from the keyboard does not count. And any gap over 45 minutes is dropped rather than billed as deliberation, which discards 406.1 hours — more than it keeps. Every one of those choices makes the figure smaller.

These are a floor on the effort, too. Verifying a build on a phone, reading away from the keyboard, and thinking between sessions all happen where no transcript can see them, and none of it is counted here. 143 sessions, last read Aug-30-26.

Estimated by work type

A later refinement of the analysis above, begun Aug-23-26 to give a clearer breakdown of what the hours were spent on. It reads a different instrument — the public work log, where a task posts a row when it opens and a row when it closes — so it starts later and it sees less.

115.2
Hours attributed to a kind of work

Since Aug-23-26

Wall-clock time with at least one task or session open. An hour with two of them open in it is still one hour.

This is deliberately not the sum of the two lines below. Adding the streams would count every overlapping hour twice, so the overlap is subtracted and shown, and the breakdown is yours to check with a calculator. The bigger number was available and was not taken.

  • Coding167 sessions75.0 hrs

    Averaging 27 min. 90% of them have both ends posted; the rest ran out of beats and were closed by the silence rule. MyStatus Work Log

  • Business67 sessions65.8 hrs

    Averaging 59 min. 34% of them have both ends posted; the rest ran out of beats and were closed by the silence rule. Biz-zy Work for MyStatus

  • Both at oncecounted once− 25.6 hrs

    Time when a coding task and a business session were open together. Real work in both columns, one hour on the clock.

What this breakdown misses, over its own dates. From Aug-23-26 onward the transcripts see 95.2 hours on the clock, and the lines above attribute 115.2 of them to a kind of work. The rest is real work that posted no row — a scheduled task, a one-question session, or the stretch between one task closing and the next opening. It is counted in the headline above and not here.

These are a floor on the effort. Verifying a build on a phone, reading, and thinking between sessions all happen off-stream, and none of it is counted here.

A session ends where the record says it ends: on a posted “Closed” or “Stopped”, or — when it simply went quiet — 30 minutes after its last beat, the same end a reconstructed stop is given. Once a stream has gone quiet long enough that the sign at the top of this page stops believing it, the session is over rather than counted through: a laptop closed overnight costs half an hour, not eight.

Estimated consulting fees

The same hours, at the rate my consulting clients actually paid, through my company, Omega Ortega. That rate is stated below, so this whole section is one multiplication you can check on a calculator.

$36,525.00
Estimated bill for the work

Since Jul-27-26

243.5 hours × $150 an hour. That is the entire calculation. Every kind of hour billed the same — coding, system design, copy, review, thinking — no adjustment, no premium for the interesting parts.

What those hours shipped is counted above: 290 features delivered.

Nobody was billed a cent of this. It is not revenue and it is not a quote. It is what the same work would have cost at human rates, put next to what it actually cost, because that comparison is the one thing this project can show that an assertion cannot.

Want this for your own project — the building, or the receipts that prove it? Head over to Omega Ortega to chat more about that.

$776.39
What it actually cost

Every bill paid, from the list above — July 2026 – August 2026.

47×
The equivalent, over the actual

How many times the real spending the rate-card figure comes to. Understated: the money runs from a longer window than the hours, and the earlier months stay in rather than being trimmed out to improve it.

$150/hr
The rate

Stated, not measured

The rate Tom’s consulting clients actually paid, through Omega Ortega. Every dollar here is this number times an hour count you can see.

Like real consulting, coding can be itemized while all the rest cannot.

The team

1
Human team members

Stated, not measured

Tom. Architecture, review, and every decision that ships.

2
AI team members

Stated, not measured

Claude Fable 5, Claude Opus 5

The split, as of today: Claude Fable 5 takes architecture, refactors, and strategy; Claude Opus 5 gets handed straight coding tasks. Everything either of them produces goes past one person who vets the logic, questions the thinking, and owns every decision that ships — which is the part that makes the first number a 1 rather than a 0.

The build

No
Coding Right Now

NO because the last thing posted was the end of a task — a "Closed" or "Stopped" entry. This flips the moment a task opens — no deploy, no delay.

Read at · Refresh

Every flip of that sign is an update on MyStatus Work Log; the entry under Last Work Task is the one it is reading.

  • Commits1,313

    Every iteration, including the wrong turns. Nothing is squashed away.

  • Features deliveredsince Aug 12, 2026290

    Finished tasks: verified, reported, merged. One merge, one task.

  • Task branches291

    All time: finished plus whatever is open right now.

  • Prompts queued15

    Work written down and waiting. This page was one of them.

  • Days since we started40

    Every day since the first commit, Jul 26, 2026, worked or not.

  • Years the idea waited6.47

    From Feb 4, 2020, when MyStatus was first written down, to the first commit.

  • Days work happened32

    Days with at least one commit. The gap between this and the row above it is the honest one.

  • Commits per day32.8

    Averaged over every day. Across only the days work happened it is 41.0 — this project runs in bursts.

  • Features per daysince Aug 12, 202612.6

    Averaged over the days in that window, not over every day since the first commit.

Counted from the repository at commit ea76d49 on , from the claude branch. These are generated by hand rather than at deploy time, and they are always exactly one commit behind — a commit cannot contain its own count.

The biz

No
Working on Biz Stuff

A session was opened and never closed, and its 30-minute beat is more than 15 minutes overdue. Reading that as NO is deliberate: silence is not evidence of work.

Read at · Refresh

Every flip of that sign is an update on Biz-zy Work for MyStatus; the entry under Last Biz Entry is the one it is reading.

Actual operating costs

Every recurring charge that keeps MyStatus alive, at the amount that actually leaves the account. No salaries — there are none — and nothing shared with the rest of Tom’s life. Beyond that, nothing is netted out and nothing is left off because it was inconvenient.

$375.69
Cost per month

Everything it takes to run this for one month, tax included.

$776.39
Spent to date

2 months on the record, July 2026 – August 2026. Kept by hand, last checked against the statements on Aug 24, 2026.

$2.68
Per feature delivered

Everything spent, divided by the 290 features delivered. Overstated: the spending runs from the first commit, the features only from Aug 12, 2026. The true figure is lower.

  • Claude subscriptionmonthly$217.8058%

    The plan that writes the code — in this company, labor is a subscription line item. Tax included: it is what leaves the account.

  • Azuremonthly$112.2030%

    The web app, the push/webhook api, and the analytics host.

  • Supabase Promonthly$35.119%

    The system of record: database, auth, storage, realtime.

  • Apple Developer Programannual$8.252%

    $99 a year, spread across the twelve months it covers.

  • my-status.app domainannual$2.331%

    $27.99 a year, spread the same way.

  • Google Play developer accountpaid once—

    Paid once, ever. Real money in the total, and no monthly figure to have.

Since Sep 2, 2026 the receipts page carries a second token figure beside the cost-weighted one: what the same tokens would cost at the AI provider’s published list rates, each exchange priced by the rate in force on the day it ran. It is not a line above — nobody is billed by the token here — but it is the number that answers what this work would cost you. What a feature costs

Conversations

Who this has been shown to and talked about with. Unlike everything else on this page these are not derived from anything — they are Tom’s count, kept by hand, last updated Aug 24, 2026. Until they are posted as dated events like the rest, they are worth exactly as much as his word.

6
Conversations held

Stated, not measured

Real conversations about MyStatus with someone outside the project.

3
Industries interviewed

Stated, not measured

Distinct fields those people worked in, counted once each.

5
Demos given

Stated, not measured

Times the product was shown actually running, rather than described.

Second conversations scheduled and pilots started belong in this group and are missing rather than zero: nobody has been keeping that count yet, which is a different statement from “it has not happened”. They appear the day the first one is recorded. Counted by hand, as of Aug 24, 2026 — the same hand, with the same caveat, as the costs above.

The service

We’re showing these numbers because we’re a new biz. We want to be upfront that this is not an at scale project yet and that’s fine because we’re very early in our experiment.

  • Tasks23

    Live tasks, public and private. Archived ones drop off.

  • Updates872

    Posted and still standing. A deleted update stops counting.

  • Owners5

    People with an account. Readers need none, and are not counted.

  • Ops tasks7

    Public MyStatus tasks that are part of our day-to-day operations.

As of . Ops tasks is the one number here you can recount yourself: every task in it is listed at /browse/ops, and what they actually do is written down.

Construction downtime

1
Recorded incident

Since Aug-06-26

10 minutes
Recorded downtime

Since Aug-06-26

We are 40 days old. The site went live Aug 6, 2026, the paid product on Sep 2, 2026. Nothing is watching it — no external monitor, no synthetic probe — so every incident here is one we caught ourselves due to our own construction. This stat is derived from MyStatus Downtime Incidents, a task served by MyStatus itself.

What is not here

No reader counts, no page views, no “most popular tracker”. Reading a MyStatus link needs no account and leaves no trace a reader would have to think about, and a leaderboard would quietly change what this product is for. There is also no “features planned” number: the roadmap lives in prose, and a hand-typed figure that drifts every week is worse than no figure at all.

Inspiration

This section exists to honor Cluetrain Manifesto’s reminders to be human, drop the corporate act and acknowledge there are no secrets. That’s why I’m building this in public and completely documenting how I dogfood the tool to build the tool.

It also honors John Wilker, my friend and old biz partner who loves Cluetrain as much as I do. He once shared how he has a dashboard of all his writing stats. It helps him write awesome booksyou should go buy! I made my own stats dashboard, which I now share with you.