Token Burn

Burn is not proof. Output is not victory. Accepted work that repeats is leverage.

today · in flight

946K active

14 sessions · spend 4.0M · Codex 479K / Claude Code 467K
by project: pal 479K · <home> 467K
top session 01a0cd8d (pal, 479K, —)

2026-09-23
as of 2026-09-23T12:55:15Z

Stupid button · 23m ago · gemma4:e4b · 2277t in / 396t out · 76.9s

Trend vs last week

Unsettled burn dropped by 4.7M (-88.1%) vs last week.
Waste increased by 1.4M (34.9%) vs last week.
Accepted burn shows no prior-window comparison.

What worked

No projects showed any accepted token burn.

What wasted tokens

The <home> project shows 0.0 acceptance ratio. Its acceptance note is "UNMEASURABLE — ran outside any git repo, so 0.0 is the meter failing to see the work, not a judgement on it," and it accounts for the 5.5M waste_active_h.
The pal project shows 0.0 acceptance ratio.
The largest waste volume is in <home> at 5.5M waste_active_h.

Why the big number is big

No agent fan-out this window.

What needs a verdict

pal + 01a0cd8d + 2026-09-23 + pal + 479K: suggested_verdict is "not_useful" because "open_loop, 479K burn — small miss".
pal + 01a07347 + 2026-09-23 + pal + 159K: suggested_verdict is "not_useful" because "open_loop, 159K burn — small miss".

One contrarian observation

The <home> project has the highest waste volume at 5.5M waste_active_h, yet its acceptance ratio is 0.0 due to its "UNMEASURABLE" status.

---

Measurability caveat (generated, not model-authored): <home> (5.5M) ran outside any git repo. There is no diff to credit, so a 0.0 acceptance ratio there is the meter failing to see the work — not a judgement on it. Only ratios on measurable: true projects are verdicts.

Rollups — acceptance lane

Headline is accepted tokens (active lane, manual label wins over git auto-classification). Unsettled = committed-not-merged + open-loop. "Active" excludes cache replay; "spend" includes it.

Last 7 days

0 accepted

unsettled 638K · acceptance 0%

spend 18.5M · 158 sessions · Claude Code 5.5M / Codex 638K

Last 30 days

0 accepted

unsettled 37.4M · acceptance 0%

spend 847.5M · 732 sessions · Claude Code 41.8M / Codex 5.3M

Last 90 days

0 accepted

unsettled 130.5M · acceptance 0%

spend 3.71B · 842 sessions · Codex 87.6M / Claude Code 52.5M

Daily heatmap — last 60 days

Each cell = one UTC day, color intensity = log10(active tokens). Hover for date + counts.

LessMore· log-scale, exact lane

Top 10 days

Date	Active	Spend	Sessions	Top project	Outcomes
2026-07-28	29.1M	1.09B	22	<home> (29.1M)	?=22
2026-08-18	21.2M	527.5M	12	<home> (21.0M)	?=14
2026-07-16	15.7M	522.1M	23	<home> (15.7M)	?=23
2026-08-06	11.1M	362.9M	27	<home> (11.1M)	?=27
2026-08-31	5.9M	144.8M	26	las (5.8M)	?=29
2026-09-08	5.1M	47.8M	28	las (4.0M)	?=28
2026-08-29	4.5M	158.4M	30	las (1.7M)	?=31
2026-08-30	4.2M	72.4M	25	las (4.2M)	?=25
2026-09-04	4.1M	90.1M	29	las (2.3M)	?=33
2026-09-11	3.9M	69.9M	30	las (3.0M)	no_repo=24 · open_loop=5 · committed=1 · PRs #21

By project — last 7 days

Project	Active
<home>	5.5M
pal	638K

By outcome — last 7 days

open_loop 2

no_repo 156

Effectiveness — last 30 days

Where delegated work shipped (top) vs. where high burn never crossed an acceptance line (bottom). Bars: accepted · unsettled · waste. Normalized to the widest project in window.

Leverage — most accepted

Project	Accepted	Ratio	Mix
pal	0	0%
las	0	0%
memory	0	0%
mainstay-av	0	0%
pal/templari	0	0%
<home>	0	0%

Waste candidates — lowest acceptance ratio (≥100K active)

Project	Accepted	Ratio	Mix
las	0	0%
pal	0	0%
<home>	0	0%
mainstay-av	0	0%
memory	0	0%

Needs validation — last 14 days (9 unsettled)

Sessions where auto-classification couldn't tell shipped work from waste. Highest-burn first. Manually label these with `label: "used"|"partial"|"redo"|"review"|"not_useful"` in token_burn.jsonl to override the auto verdict.

Date	Project	Session	Active	Auto	cwd
2026-09-11	las	4b15a997	3.0M	committed · #21 (OPEN)	.../las/.worktrees/mvp
2026-09-16	las	4b15a997	1.6M	committed · #21 (OPEN)	.../las/.worktrees/mvp
2026-09-23	pal	01a0cd8d	479K	open_loop	.../pal
2026-09-11	pal	01a07347	448K	open_loop	.../pal
2026-09-11	pal	01a0925e	165K	open_loop	.../pal
2026-09-23	pal	01a07347	159K	open_loop	.../pal
2026-09-11	pal	01a0925f	151K	open_loop	.../pal
2026-09-11	pal	01a0925e	17K	open_loop	.../pal
2026-09-11	pal	01a0888c	17K	open_loop	.../pal

Recent events — last 30 rows

Latest enriched rows from the JSONL (sorted by date desc, active desc as tiebreaker). M = manually labeled.

Date	Tool	Project	Session	Active	Output	Outcome
2026-09-23	Codex	pal	01a0cd8d	479K	3K	open_loop
2026-09-23	Codex	pal	01a07347	159K	17K	open_loop
2026-09-23	CC	<home>	b3aac828	36K	924	no_repo
2026-09-23	CC	<home>	6d0125de	36K	826	no_repo
2026-09-23	CC	<home>	2f8f088d	36K	852	no_repo
2026-09-23	CC	<home>	5d6c082d	36K	768	no_repo
2026-09-23	CC	<home>	1fb58060	36K	760	no_repo
2026-09-23	CC	<home>	bb1f34a4	36K	728	no_repo
2026-09-23	CC	<home>	c05b66d8	36K	514	no_repo
2026-09-23	CC	<home>	d4765b66	36K	612	no_repo
2026-09-23	CC	<home>	3602235d	36K	580	no_repo
2026-09-23	CC	<home>	449d0f0d	36K	554	no_repo
2026-09-23	CC	<home>	9b67c44b	36K	536	no_repo
2026-09-23	CC	<home>	45da7789	36K	516	no_repo
2026-09-22	CC	<home>	99d610e2	36K	990	no_repo
2026-09-22	CC	<home>	c7d27122	36K	978	no_repo
2026-09-22	CC	<home>	ec5a20c2	36K	968	no_repo
2026-09-22	CC	<home>	ec986243	36K	900	no_repo
2026-09-22	CC	<home>	54704d28	36K	928	no_repo
2026-09-22	CC	<home>	7670347a	36K	826	no_repo
2026-09-22	CC	<home>	d24c5329	36K	726	no_repo
2026-09-22	CC	<home>	c1c2e6f8	36K	724	no_repo
2026-09-22	CC	<home>	e5f1a06f	36K	716	no_repo
2026-09-22	CC	<home>	4aaec215	36K	994	no_repo
2026-09-22	CC	<home>	77d8af60	36K	672	no_repo
2026-09-22	CC	<home>	c5b98857	36K	748	no_repo
2026-09-22	CC	<home>	73a85ea6	36K	738	no_repo
2026-09-22	CC	<home>	cc4aa57d	36K	602	no_repo
2026-09-22	CC	<home>	1c5f110c	36K	674	no_repo
2026-09-22	CC	<home>	9df3f079	36K	652	no_repo

Generated 2026-09-23 12:55 UTC from ~/ideas/pal/dashboards/token_burn.jsonl (911 rows). Sources: ~/.codex/sessions/ (exact), ~/.claude/projects/-home-khoward/ (exact). No estimate lane currently active.