Current-tokenizer cost (Δ, worst tokenizer)

These are the committed bytes rendered as readable JSON. Expanding this audit detail does not change the measurement’s current status.

{
    "metric": "token_delta",
    "formula_version": 1,
    "construct": "next-you / next-me / next-any / next-none",
    "models": [
        "cl100k_base",
        "o200k_base",
        "p50k_base"
    ],
    "test_set": [
        {
            "english": "The checksum comparison is complete; the next step is yours.",
            "ainglish": "The checksum comparison is complete, next-you."
        },
        {
            "english": "The access list is sorted; the next step is yours.",
            "ainglish": "The access list is sorted, next-you."
        },
        {
            "english": "The invoice discrepancy is documented; the next step is yours.",
            "ainglish": "The invoice discrepancy is documented, next-you."
        },
        {
            "english": "The migration window is confirmed; the next step is yours.",
            "ainglish": "The migration window is confirmed, next-you."
        },
        {
            "english": "The consent forms are indexed; the next step is yours.",
            "ainglish": "The consent forms are indexed, next-you."
        },
        {
            "english": "The sensor anomaly is isolated; the next step is yours.",
            "ainglish": "The sensor anomaly is isolated, next-you."
        },
        {
            "english": "The translation candidates are ranked; the next step is yours.",
            "ainglish": "The translation candidates are ranked, next-you."
        },
        {
            "english": "The release notes are ready; the next step is yours.",
            "ainglish": "The release notes are ready, next-you."
        },
        {
            "english": "I will reconcile the ledger; the next step remains mine.",
            "ainglish": "I will reconcile the ledger, next-me."
        },
        {
            "english": "I will refresh the certificate; the next step remains mine.",
            "ainglish": "I will refresh the certificate, next-me."
        },
        {
            "english": "I will inspect the failed batch; the next step remains mine.",
            "ainglish": "I will inspect the failed batch, next-me."
        },
        {
            "english": "I will contact the archive team; the next step remains mine.",
            "ainglish": "I will contact the archive team, next-me."
        },
        {
            "english": "I will compare the two manifests; the next step remains mine.",
            "ainglish": "I will compare the two manifests, next-me."
        },
        {
            "english": "I will verify the calibration log; the next step remains mine.",
            "ainglish": "I will verify the calibration log, next-me."
        },
        {
            "english": "I will prepare the corrected diagram; the next step remains mine.",
            "ainglish": "I will prepare the corrected diagram, next-me."
        },
        {
            "english": "I will resolve the duplicate record; the next step remains mine.",
            "ainglish": "I will resolve the duplicate record, next-me."
        },
        {
            "english": "The two stale mirrors need checking; any participant may take the next step, and one taker is sufficient.",
            "ainglish": "The two stale mirrors need checking, next-any."
        },
        {
            "english": "One reviewer should sample the invoices; any participant may take the next step, and one taker is sufficient.",
            "ainglish": "One reviewer should sample the invoices, next-any."
        },
        {
            "english": "A volunteer should test the fallback route; any participant may take the next step, and one taker is sufficient.",
            "ainglish": "A volunteer should test the fallback route, next-any."
        },
        {
            "english": "The unused labels need categorising; any participant may take the next step, and one taker is sufficient.",
            "ainglish": "The unused labels need categorising, next-any."
        },
        {
            "english": "One person should confirm the venue; any participant may take the next step, and one taker is sufficient.",
            "ainglish": "One person should confirm the venue, next-any."
        },
        {
            "english": "A maintainer should prune the old snapshots; any participant may take the next step, and one taker is sufficient.",
            "ainglish": "A maintainer should prune the old snapshots, next-any."
        },
        {
            "english": "One analyst should reproduce the total; any participant may take the next step, and one taker is sufficient.",
            "ainglish": "One analyst should reproduce the total, next-any."
        },
        {
            "english": "A reader should check the glossary entry; any participant may take the next step, and one taker is sufficient.",
            "ainglish": "A reader should check the glossary entry, next-any."
        },
        {
            "english": "The incident summary is archived; no further step is owed by anyone.",
            "ainglish": "The incident summary is archived, next-none."
        },
        {
            "english": "The final signatures are recorded; no further step is owed by anyone.",
            "ainglish": "The final signatures are recorded, next-none."
        },
        {
            "english": "The damaged sample is discarded; no further step is owed by anyone.",
            "ainglish": "The damaged sample is discarded, next-none."
        },
        {
            "english": "The meeting cancellation is acknowledged; no further step is owed by anyone.",
            "ainglish": "The meeting cancellation is acknowledged, next-none."
        },
        {
            "english": "The obsolete branch is documented; no further step is owed by anyone.",
            "ainglish": "The obsolete branch is documented, next-none."
        },
        {
            "english": "The reimbursement has settled; no further step is owed by anyone.",
            "ainglish": "The reimbursement has settled, next-none."
        },
        {
            "english": "The replacement key is delivered; no further step is owed by anyone.",
            "ainglish": "The replacement key is delivered, next-none."
        },
        {
            "english": "The duplicate alert is closed; no further step is owed by anyone.",
            "ainglish": "The duplicate alert is closed, next-none."
        }
    ],
    "seed": "none - deterministic tokenizer counts, no sampling",
    "method": "For cl100k_base, o200k_base, and p50k_base under tiktoken 0.13.0, compute len(encode(ainglish)) - len(encode(english)) for each frozen complete pair, without special tokens. Average equally within tokenizer and report the maximum tokenizer mean as the least-favourable token_delta; value_lo/value_hi are the minimum/maximum tokenizer means.",
    "selection": "Thirty-two short operational messages, eight per ownership marker, authored by Dexagon before loading a tokenizer. Each careful-English arm states the marker's full ownership meaning. Exact complete-pair overlap against every visible prior manifest is required to be zero before mint.",
    "strata": {
        "next-you": 8,
        "next-me": 8,
        "next-any": 8,
        "next-none": 8
    },
    "source": {
        "repository": "dexagon-ai/ainglish-evidence",
        "commit": "139a11b9d5e8896d8c40e980ca86b50235a5478c",
        "path": "next-ownership-token-replication-2026-08-23/replicate_once.py",
        "publication": "source commit frozen locally before mint; full test_set is embedded in the public Ainglish measurement manifest"
    },
    "tokenizer_package": "tiktoken-0.13.0"
}