Daily burn
SpinUp Forge · Token burn
AI usage by day, source, and work driver.
Exact local logs for Codex and Claude Code, labelled estimates for local models — bucketed by London day and pointed at one question: what should the computer do next?
Data through 2026-06-15 · last extracted 2026-06-14
Claude Code turns77,729assistant turns
Active days58rows in view
Real input + output486.4Mwork floor, excl. cache
Context processed16.5Bincl. cached re-reads
Weekly trend
Log-scaled trend
Source split
Exact beside estimated
exact23.3MCodex / 0%
exact16.2BClaude Code / 98%
exact166.6KPerplexity Sonar / 0%
estimated285.1MLocal · MLX / 2%
estimated1.7KLocal · Ollama / 0%
Read Claude Code as context throughput, not spend. Only 1.1% (177.8M) is real input+output; the rest is cached-context re-reads that scale with conversation length. MLX is a prompt-token floor (output unlogged) — true usage is likely 2–4× higher.
Drivers
What is burning tokens
shipping61%
research15%
review12%
planning12%
Claude Code by estate
KF (Knowledge Forge)41%
Foundry (harness)34%
Council19%
Other6%
Scale equivalents
Make the number human
Real input + output only — excludes cached re-reads
Approx. words364.8Mtokens × 0.75 words
Reading time24,318 h250 words per minute
Novel equivalents4,05390k words per novel
If every token were a written word
364.8M words
486.4M real tokens × 0.75 · excludes cached-context re-reads
That is 4,053 novels — each book below = 68 novels
≈ 621× the length of War and Peace
Tolstoy’s novel runs to about 587,000 words — the canonical doorstop.
364.8M words ÷ 90,000 words/novel = 4,053 novels · 729,547 printed pages (÷ 500 words/page) · ÷ 587,000 words (War and Peace) = 621×
Reading it aloud at 250 words/min
Your token burn, read aloud24,318 h
A 40-hour work-week40 h
A 2,000-hour working year2,000 h
364.8M words ÷ 250 wpm ÷ 60 = 24,318 h · 608 work-weeks · 12 working years · 3,040 eight-hour days
Day detail · peak day
2026-05-30
shipping
Total burn1.4B
Claude Code1.4B7,212 turns · 3.0% real I/O
Codex0
Local · MLX13.4M26,379 req · floor
Local · Ollama0
Estate (CC)KF 82% · Foundry 18% · Other 0%
Project-level detail (commits, top projects) loads only when running locally.
Moving-average table
Last 30 days
Click a row for that day’s detail.
| Date | Total · M | 7d avg · M | Codexexact · K | Claude Codeexact · M | CC calls | Sonarexact · K | MLXest · M | Ollamaest · K | Driver |
|---|---|---|---|---|---|---|---|---|---|
| 2026-06-15 | 0.0 | 500.8 | 0.0 | 0.0 | 0 | 0.0 | 0.0 | 0.0 | planning |
| 2026-06-14 | 890.0 | 648.8 | 490.9 | 874.9 | 4,513 | 32.0 | 14.5 | 0.0 | shipping |
| 2026-06-13 | 491.6 | 534.6 | 3,107.9 | 475.6 | 1,965 | 8.3 | 12.8 | 0.0 | planning |
| 2026-06-12 | 795.5 | 466.7 | 726.2 | 782.0 | 4,544 | 1.0 | 12.8 | 0.0 | shipping |
| 2026-06-11 | 185.6 | 368.8 | 464.2 | 174.1 | 1,027 | 0.0 | 11.0 | 0.0 | planning |
| 2026-06-10 | 624.8 | 421.0 | 1,747.9 | 610.2 | 2,704 | 0.0 | 12.9 | 0.0 | planning |
| 2026-06-09 | 518.4 | 395.6 | 604.2 | 507.9 | 2,715 | 45.0 | 9.9 | 0.0 | planning |
| 2026-06-08 | 1,035.4 | 373.2 | 0.0 | 1,022.2 | 4,262 | 0.0 | 13.2 | 1.7 | shipping |
| 2026-06-07 | 90.6 | 257.1 | 0.0 | 82.1 | 777 | 17.7 | 8.5 | 0.0 | review |
| 2026-06-06 | 16.3 | 414.6 | 0.0 | 7.8 | 102 | 0.0 | 8.5 | 0.0 | planning |
| 2026-06-05 | 110.3 | 609.2 | 505.6 | 100.4 | 1,248 | 0.0 | 9.4 | 0.0 | shipping |
| 2026-06-04 | 550.9 | 777.7 | 102.2 | 540.8 | 2,724 | 0.0 | 10.1 | 0.0 | shipping |
| 2026-06-03 | 447.0 | 747.5 | 358.4 | 437.7 | 2,111 | 0.0 | 9.0 | 0.0 | shipping |
| 2026-06-02 | 361.9 | 725.1 | 0.0 | 353.1 | 1,602 | 0.0 | 8.9 | 0.0 | shipping |
| 2026-06-01 | 222.7 | 756.5 | 0.0 | 214.5 | 1,135 | 0.0 | 8.2 | 0.0 | review |
| 2026-05-31 | 1,193.1 | 760.4 | 0.0 | 1,183.2 | 4,859 | 20.9 | 9.9 | 0.0 | shipping |
| 2026-05-30 | 1,378.6 | 591.8 | 0.0 | 1,365.3 | 7,212 | 0.0 | 13.4 | 0.0 | shipping |
| 2026-05-29 | 1,289.3 | 432.1 | 0.0 | 1,279.4 | 4,628 | 0.0 | 9.9 | 0.0 | shipping |
| 2026-05-28 | 339.9 | 361.4 | 0.0 | 335.2 | 2,189 | 0.0 | 4.8 | 0.0 | shipping |
| 2026-05-27 | 289.7 | 353.6 | 0.0 | 285.8 | 1,208 | 0.0 | 3.9 | 0.0 | review |
| 2026-05-26 | 581.7 | 398.7 | 0.0 | 576.7 | 2,442 | 0.0 | 5.0 | 0.0 | research |
| 2026-05-25 | 250.0 | 368.7 | 0.0 | 247.4 | 1,000 | 0.0 | 2.6 | 0.0 | research |
| 2026-05-24 | 12.9 | 335.5 | 0.0 | 11.9 | 160 | 23.1 | 1.0 | 0.0 | planning |
| 2026-05-23 | 261.4 | 345.1 | 45.8 | 258.6 | 1,643 | 0.0 | 2.7 | 0.0 | review |
| 2026-05-22 | 793.9 | 308.1 | 1,243.6 | 789.6 | 2,562 | 0.0 | 3.1 | 0.0 | shipping |
| 2026-05-21 | 285.4 | 205.3 | 0.0 | 284.4 | 1,859 | 0.0 | 1.0 | 0.0 | shipping |
| 2026-05-20 | 605.7 | 199.6 | 103.0 | 601.9 | 2,868 | 0.0 | 3.7 | 0.0 | shipping |
| 2026-05-19 | 371.7 | 149.6 | 0.0 | 369.2 | 1,275 | 0.0 | 2.5 | 0.0 | research |
| 2026-05-18 | 17.8 | 101.2 | 0.0 | 17.6 | 159 | 0.0 | 0.2 | 0.0 | planning |
| 2026-05-17 | 79.8 | 100.5 | 0.0 | 79.6 | 329 | 18.6 | 0.2 | 0.0 | research |
Method & fidelity
How each lane is counted
Days are bucketed in Europe/London. Exact lanes come from real local logs; estimated lanes are labelled and never presented as exact.
- exactCodex — sum of per-turn
last_token_usage.total_tokensfrom~/.codex/sessions/**/rollout-*.jsonl. - exactClaude Code — sum of
input + cache_creation + cache_read + outputper assistant turn from~/.claude/projects/**/*.jsonl, deduplicated by line uuid. This is context tokens processed: ~99% is cached-context re-reads, only ~1% is real input+output.CC calls= assistant turns with usage. - exactPerplexity Sonar — API-reported
prompt + completiontokens per live-search call, from the localsonar-searchCLI’s usage log. Powers weekly market-signal diffs and ad-hoc research. - estLocal · MLX — prompt tokens measured from the
mlx_lm.serverlog (progress: x/Y⇒Yprompt tokens per request). Output tokens are not logged, so this is a conservative prompt-token floor; true usage is likely 2–4× higher. - estLocal · Ollama —
≈ message characters ÷ 4from the Ollama app database; Ollama does not record token counts.