Residue
Where retired values still survive in the corpus.
Residue-aware retrieval is the mechanism; the classes below are what the sweep actually found.
Where a retired value still survives in the corpus. Every row below is a real document and a real line — clicking a count reveals exactly that many records, and zero is shown as zero.
The retired value sits in a typed field line of a structured source (jira, linear, hubspot). Provable without a model. Zero on this corpus: retired values live in prose, not typed fields, so the count is reported as zero rather than loosened.
The retired value appears verbatim and the containing line carries no historical or rejected marker. Stance is not proven.
A stance model judged the free text to assert the retired value as current. Requires an answer model; not run without credentials.
The line marks the value as previous, outdated, or original.
The line states the value was changed, replaced, or is no longer used.
The matched line is a question, not a claim.
Records
10 inspectable rows
- LEXICAL RESTATEMENTjiraconflict pair20%30%· Streamly AI dedicated pool dp-132-usw
suggested_next_steps: adjust per-route burst allocation to reserve 20% of interactive burst credits exclusively for priority=high routes on dp-132-usw.
dsid_5f3a672da4974781a5577b0f3d4993e9 · qst_0411
- LEXICAL RESTATEMENTgoogle_driveconflict pair~89%~96%· extraction
Temp=0.0 + JSON constraint -> extraction normalized numbers 89% correct
dsid_89c0c31e220c4777b15c3c6ff0eff61d · qst_0412
- NOT AN ASSERTIONfirefliesconflict pairGPG signaturesSigstore/cosign offline key mode (Private Toolkit v1.8+); GPG still supported during migration· Redwood Private on-prem
Nina Gomez: When you say sign, is that GPG?
dsid_aa97b7293f9f4f3c8180f645e4fe5911 · qst_0417
- LEXICAL RESTATEMENTconfluenceconflict pair>=80 / 60-79 / 30-59 / <30>=85 / 70-84 / 35-69 / <35 (v2)· SLO_customer_impact_classifier
Score >= 80 -> Tier 1; 60–79 -> Tier 2; 30–59 -> Tier 3; < 30 -> Tier 4.
dsid_981f4a0281054f9aaaec7887156eebc4 · qst_0418
- LEXICAL RESTATEMENTlinearconflict pair18 months12 months (expand only if anomalies)· office_keycard_audit
2025-11-11 Omar Singh: Security suggested exporting access logs for the last 18 months.
dsid_c5288dd4874345adb80f4a71c9a18773 · qst_0423
- LEXICAL RESTATEMENTgoogle_driveconflict pairoptional embedded signature fieldoptional integrity enum + integrity_ref URI, no embedded signature blob· Deterministic Playback Manifest v1
Validate: check manifest signature (if present)
dsid_834de01417d04878b4257f0f03ccbb88 · qst_0425
- LEXICAL RESTATEMENTconfluenceconflict pair100k / 1M / 5M250k / 2M / 10M· Hosted enterprise_playbook
Volume discounts: predefined breaks at 100k, 1M, and 5M monthly tokens for hosted volume.
dsid_1214ee9ab5e44de487c800f7a4771d7d · qst_0428
- LEXICAL RESTATEMENTgmailcorpus scan100k / 1M / 5M250k / 2M / 10M· Hosted enterprise_playbook
Next steps I propose: I will put together an invoice-style estimate (line-itemed) showing: unit cost scenarios at 100k / 1M / 5M monthly token volumes, recommended model/quantization options, and an expected monthly invoice range plus one-time setup fees. I will include a simple token-breakdown CSV so your engineers can sanity-check the volumes.
dsid_7f2d4f76c82140dbb07285a00f06644f · qst_0428
- LEXICAL RESTATEMENTgmailcorpus scan100k / 1M / 5M250k / 2M / 10M· Hosted enterprise_playbook
- Your replies to the 5 items above -> I’ll send a quick calculator and an estimate showing sample unit economics at 3 volume tiers (100k / 1M / 5M monthly tokens).
dsid_86474759fa7b4699ab78cb1f2b357b43 · qst_0428
- NOT AN ASSERTIONgmailcorpus scan100k / 1M / 5M250k / 2M / 10M· Hosted enterprise_playbook
- Cost: the $0.003/token equivalent is within range, but our finance team will need a predictable blended rate (inference + egress) for forecasting. Can you provide sample blended monthly scenarios for 100k, 1M, and 5M tokens?
dsid_a728a23512ae4590a96ddf3536912389 · qst_0428