OGCOM TICKS

Bottom line

Loading telemetry…

—

TICKS SCORE —/100

Want this read on a different session? Drop its transcript file here — a .jsonl from %USERPROFILE%\.claude\projects\. That format also contains your own messages, so it can judge your prompting; a CSV can only judge the cost. Nothing leaves your browser.

Where to break the session

Not a rule of thumb — a simulation run on this session's own request-by-request trajectory. It replays the same work split across a different number of sessions, charging the opening load each time a new one starts. The model reproduces the measured single-session total exactly, which is what makes the other rows trustworthy.

Sessions	Requests each	Weighted tokens	Change	Ends near

Account monitor

Session	Started	Hrs	Reqs	New material	Re-reads	Output	Out/hr	Est. cost

Are you maximising the weekly limit?

Your weekly allowance is not published as a token number, so this cannot be guessed — it has to be calibrated from your own account. Open Settings › Usage in Claude, read the weekly progress bar, and type the percentage here. Everything below then becomes exact for you.

Your plan

Weekly bar reads now Settings › Usage, weekly bar, right now. Cumulative for the week.

…and before this session Only the difference can price a session. Without it, this shows what is left, not what today cost.

Do cache re-reads count? Anthropic states cached project content does not count when reused. Whether that extends to session caching is not documented.

What is actually happening

Read this first. Everything below it is the evidence.

Your five questions, answered from your own data

Each one is measured against the session loaded above — not against a rule of thumb.

Things you are not doing that you should be

Checked against this session. A filled mark means the habit is already in place; an open mark means it is available to you and unused.

Where the money actually went

Not tokens — dollars. Only the gold bar is work you asked for. Everything above it is the cost of carrying the conversation: storing what was said, then re-sending it on every request that follows.

Cache read · 0.10× Fresh input · 1.00× Cache write 5m · 1.25× Cache write 1h · 2.00× Output

Context accumulation

How heavy the conversation got. Each dot is one request, and its height is everything that had to be re-sent to make it. The line only ever goes up — that is the whole problem in one picture.

Request ledger

The raw evidence, one row per billed request, if you want to check any number above.

Time	Model	Fresh	Cache read	Write 1h	Output	Thinking	Total in	I/O	Hit %	Cost

What these dollar figures are. They are list API prices, calculated from the published rate card as of 2026-08-25 — not a bill. A subscription seat is not charged per token, so read every figure here as a way to compare one working habit against another, not as money owed. Token counts are exact, taken from the session's own records. Your typed tokens are estimated from character count at roughly four characters per token.