Measurement · not a scoreboard

Token Burn

I run work across Claude, Codex, ChatGPT and local agents. Token usage isn't a leaderboard — it's a feedback loop for delegated intelligence: what work actually moved, whether the result got used, and whether the review burden was worth it. The hard part is honesty — you can't trust one token number across tools that count differently, so I keep three fidelity lanesseparate and never sum them into a single figure.

The three lanes (kept apart)

Exact tokens

89.9M

tokens · 78 receipts

Counted by the provider. Hard numbers.

Measured activity

22K

ops · 20 receipts

Real ops (tool calls, runs) — not tokens.

Estimate band

40K–90K

tokens · 1 receipts

Best-guess ranges. Never precise.

Three numbers, three confidence levels. Adding them would invent precision that doesn't exist.

Usage over time · daily exact-token total

2026-05-11tokens/day · exact lane · log-scaled · peak 10.2M2026-06-10

Context-mode savings — bytes the model did not have to read

Context-mode receipt

147.3 MB

avoided from model context across filtered rows

147.3 MB raw inspected/indexed/processed → 0 B returned (100% avoided). This is the dashboard's “don't make the model read the haystack” lane.

Savings receipts come from context-mode style work: indexing, searching, sandboxed command/file processing, and returning only matched windows. They are not token totals; they are evidence of avoided context stuffing.

Context Hygiene≈36.8Mtokens saved

Where the work runs · by tool

Hermes default44

Claude Code40

Gideon/Ollama9

Codex5

ChatGPT1

bars = receipt count per tool (any lane)

Exact-token weight · by model

claude-opus-4-841.5M

claude-opus-4-728.5M

OpenAI Codex19.6M

anthropic226K

gideon:latest14K

exact lane only · measured provider tokens

Per-agent burn · exact-token activity

Total

89.9M tok

Claude Code

70.1M tok

Hermes default

19.8M tok

Gideon/Ollama

14K tok

lessmore · exact-token lane · shared log scale across agents

2026-05-11 → 2026-06-10 · Total plus one band per tool · log-scaled per day

Receipts · scrubbed

Date	Tool	Model	Category	Lane	Value	Outcome	Next move
2026-06-10	Codex	activity-only	operations	activity	270	used	shrink-context
2026-06-09	Gideon/Ollama	gideon:latest	automation	exact	7.5K	used	repeat
2026-06-09	Codex	activity-only	operations	activity	270	used	shrink-context
2026-06-09	Claude Code	context-mode	operations	activity	2.0K	used	shrink-context
2026-06-09	Claude Code	context-mode	operations	activity	2.1K	used	shrink-context
2026-06-09	Claude Code	claude-opus-4-8	agent-coding	exact	5.2M	used
2026-06-09	Hermes default	OpenAI Codex	agent-runtime	exact	355K	used	monitor
2026-06-09	Hermes default	OpenAI Codex	agent-runtime	exact	23K	used	monitor
2026-06-09	Hermes default	OpenAI Codex	agent-runtime	exact	783K	used	monitor
2026-06-08	Gideon/Ollama	gideon:latest	automation	exact	5.8K	used	repeat
2026-06-08	Codex	activity-only	operations	activity	270	used	shrink-context
2026-06-08	Claude Code	context-mode	operations	activity	2.0K	used	shrink-context
2026-06-08	Claude Code	context-mode	operations	activity	2.0K	used	shrink-context
2026-06-08	Claude Code	claude-opus-4-8	agent-coding	exact	1.1M	used
2026-06-08	Hermes default	OpenAI Codex	agent-runtime	exact	1.0M	used	monitor
2026-06-08	Hermes default	OpenAI Codex	agent-runtime	exact	497K	used	monitor
2026-06-08	Hermes default	OpenAI Codex	agent-runtime	exact	331K	used	monitor
2026-06-07	Gideon/Ollama	gideon:latest	automation	exact	731	used	repeat
2026-06-07	Gideon/Ollama	gideon:latest	automation	exact	325	used	repeat
2026-06-07	Codex	activity-only	operations	activity	270	used	shrink-context
2026-06-07	Claude Code	context-mode	operations	activity	2.0K	used	shrink-context
2026-06-07	Hermes default	OpenAI Codex	operations	activity	0	used	shrink-context
2026-06-07	Claude Code	claude-opus-4-8	agent-coding	exact	7.4M	used
2026-06-07	Hermes default	OpenAI Codex	agent-runtime	exact	760K	used	monitor
2026-06-07	Hermes default	OpenAI Codex	agent-runtime	exact	436K	used	monitor
2026-06-07	Hermes default	OpenAI Codex	agent-runtime	exact	535K	used	monitor
2026-06-06	Claude Code	context-mode	operations	activity	1.9K	used	shrink-context
2026-06-06	Codex	activity-only	operations	activity	270	used	shrink-context
2026-06-06	Hermes default	OpenAI Codex	operations	activity	14	used	shrink-context
2026-06-06	Claude Code	context-mode	operations	activity	1.9K	used	shrink-context
2026-06-06	Claude Code	context-mode	operations	activity	1.9K	used	shrink-context
2026-06-06	Claude Code	context-mode	operations	activity	2.0K	used	shrink-context
2026-06-06	Claude Code	claude-opus-4-8	agent-coding	exact	5.9M	used
2026-06-06	Hermes default	OpenAI Codex	agent-runtime	exact	822K	used	monitor
2026-06-06	Hermes default	OpenAI Codex	agent-runtime	exact	458K	used	monitor
2026-06-06	Hermes default	OpenAI Codex	agent-runtime	exact	83K	used	monitor
2026-06-05	Gideon/Ollama	gideon:latest	verification	exact	33	used	repeat
2026-06-05	Gideon/Ollama	gideon:latest	verification	exact	33	used	repeat
2026-06-05	Claude Code	context-mode	operations	activity	843	used	shrink-context
2026-06-05	Claude Code	claude-opus-4-8	agent-coding	exact	1.5M	used
2026-06-05	Hermes default	OpenAI Codex	agent-runtime	exact	1.3M	used	monitor
2026-06-04	Claude Code	claude-opus-4-8	agent-coding	exact	1.1M	used
2026-06-04	Hermes default	OpenAI Codex	agent-runtime	exact	648K	used	monitor
2026-06-04	Hermes default	OpenAI Codex	agent-runtime	exact	67K	used	monitor
2026-06-03	Claude Code	claude-opus-4-8	agent-coding	exact	2.8M	used
2026-06-03	Hermes default	OpenAI Codex	agent-runtime	exact	948K	used	monitor
2026-06-03	Hermes default	OpenAI Codex	agent-runtime	exact	254K	used	monitor
2026-06-02	Claude Code	claude-opus-4-8	agent-coding	exact	4.2M	used
2026-06-02	Hermes default	OpenAI Codex	agent-runtime	exact	898K	used	monitor
2026-06-02	Hermes default	OpenAI Codex	agent-runtime	exact	7.8K	used	monitor
2026-06-02	Hermes default	OpenAI Codex	agent-runtime	exact	110K	used	monitor
2026-06-01	Gideon/Ollama	gemma4:26b-a4b	automation	activity	1.2K	used	repeat
2026-06-01	Claude Code	claude-opus-4-8	agent-coding	exact	519K	used
2026-06-01	Hermes default	OpenAI Codex	agent-runtime	exact	1.3M	used	monitor
2026-06-01	Hermes default	OpenAI Codex	agent-runtime	exact	19K	used	monitor
2026-06-01	Hermes default	OpenAI Codex	agent-runtime	exact	143K	used	monitor
2026-05-31	Claude Code	claude-opus-4-8	agent-coding	exact	617K	used
2026-05-31	Hermes default	OpenAI Codex	agent-runtime	exact	655K	used	monitor
2026-05-31	Hermes default	OpenAI Codex	agent-runtime	exact	397K	used	monitor
2026-05-30	Claude Code	claude-opus-4-8	agent-coding	exact	9.4M	used

Project, client and job detail are stripped at build time — only tool, model, generic category and outcome are shown. Snapshot generated 6/10/2026 · 99 source receipts.