Chapter 4.5 follows chapter 4.4, Physical differences between the rungs, and precedes chapter 4.6, Application of a profile to a build, within part 4, the profile library.
The measured size of documentation is described on R3. The measurement policy compares documentation with its source code through tokens. The billed-token model is the recorded model because it measures the units charged to the agent. One shared denominator keeps each comparison consistent. source: 3. LAP Profile Library/measurement_policy.md line 22
4.5.1 Token models of the documentation profiles
The policy defines three token models. The application programming interface (API) reported count uses the count returned by the API or command-line usage record. The characters-divided-by-four estimator derives a token estimate from character count. The words-divided-by-zero-point-seven-five estimator derives a legacy estimate from word count. A comparison uses one model for both values. The billed-token model is the canonical model when it is available. The character estimator remains an estimate, and the word estimator is reported only beside it.
source: 3. LAP Profile Library/measurement_policy.md line 22
4.5.2 Rung payload budgets of the documentation profiles
Each profile has a payload budget file. The budget records a target and the applicable structural limits for that profile. Recorded payload uses the billed-token model when the measurement is available. Historical estimates retain their recorded model, while a new comparison identifies its model.
L1: 678 tokens and 0.228 percent source: Appendix B, Normative Reference.md (Plan document 3), section (c), table at line 97
L1H: 4,399 tokens and 1.479 percent source: Appendix B, Normative Reference.md (Plan document 3), section (c), table at line 98
L2C: 5,782 tokens and 1.944 percent source: Appendix B, Normative Reference.md (Plan document 3), section (c), table at line 99
L2S: 8,018 tokens and 2.696 percent source: Appendix B, Normative Reference.md (Plan document 3), section (c), table at line 100
L2T: 10,749 tokens and 3.615 percent source: Appendix B, Normative Reference.md (Plan document 3), section (c), table at line 101
L3B: 11,876 tokens and 3.994 percent source: Appendix B, Normative Reference.md (Plan document 3), section (c), table at line 102
L4P: 25,057 tokens and 8.427 percent source: Appendix B, Normative Reference.md (Plan document 3), section (c), table at line 103
L5P: 43,683 tokens and 14.690 percent source: Appendix B, Normative Reference.md (Plan document 3), section (c), table at line 104
4.5.3 Canonical measurement of the documentation profiles
The canonical measurement was made on 2026-08-06. The measurement file records the measurement identifier, date, token model, estimate, policy, method, source basis, and rung results. It records the billed-token model as the figure of record for the shared source basis.
The method uses a native tokenizer, one shared denominator, and scaffolding overhead subtracted before the payload is recorded. source: api_reported_measurement_2026-08-06.json key method
The character-per-token spread runs from 3.09 to 4.44. source: Appendix B, Normative Reference.md line 106
4.5.4 Profile instrumentation of the documentation profiles
The profile instrumentation program prepares a clean code tree, selects a profile from the registry, copies the tree to an output location, and composes the instrumentation prompt. It invokes the command-line agent, runs the conformance gate, checks code identity and required artifacts, and records the resulting payload. A failed gate ends the run with failure; a passed gate records the profile measurement.
source: 1. Harness/scripts/instrument_profile.py
Figure D-L4-1. The three token models and the one recorded as the figure of record
%% figure D-L4-1 flowchart TB subgraph g1[" "] n1_1["chars_div4"] --> n1_2["words_div0_75"] n1_2 --> n1_3["api_reported"] n1_3 --> n1_4["measurement_policy.md"] end subgraph g2[" "] n2_1["lap_profile_index.csv"] --> n2_2["payload_pct_chars"] n2_2 --> n2_3["payload_pct_words"] n2_3 --> n2_4["step"] end subgraph g3[" "] n3["method"] end n1_4 --> n2_1 n2_4 --> n3 n2_4 -- "api_reported_measurement_2026-08-06.json" --> n2_4_reference["reference"]
source: operations/site-ia/A3-diagram-specification.md
source: 1. Harness/scripts/instrument_profile.py
The profile instrumentation process from source tree preparation through measurement.
Figure D-L4-2. A chart comparing the recorded payload of each profile from the measurement file with the profile index, using a shared scale and identifying the billed-token model.
%% figure D-L4-2 flowchart LR subgraph g1[" "] n1["A chart comparing the recorded payload of each profile from the measurement file with the profile index<br/>using a shared scale and identifying the billed-token model"] end
source: operations/site-ia/A3-diagram-specification.md
source: 3. LAP Profile Library/api_reported_measurement_2026-08-06.json key rungs
The measured payload comparison across the profile index.