docs+marketing: correct uncovered-codes count + tighten claim phrasing#18
Merged
Merged
Conversation
…im phrasing
GPT external review of the v3 bench results spotted two issues:
1. The "honest gaps" section listed 6 codes but said "5 of 21 codes
unexercised". Actual count is 8 codes uncovered, 13 exercised
(verified by grepping EXPECTED_ISSUE_CODES against the
ConveyancingIssueCode union). The list was also missing
BOUNDARY_DISPUTE_UNRESOLVED (currently tested only as a must-not-
emit guard, not legitimately emitted by any scenario) and
PLANNING_TA6_SEARCH_DISCREPANCY (no scenario lists it as expected).
Fixed: BASELINES.md now records "13 of 21 exercised, 8 uncovered"
with the full uncovered list and notes the speculative-priority
order for future growth (TITLE_DEFECTIVE_INDEMNITY_REQUIRED,
SEARCH_CONTAMINATED_LAND, MORTGAGE_OFFER_EXPIRY_RISK, LEASE_
ALTERATIONS_CONSENT - all common lender / client concerns).
2. Marketing claim was missing the "internal" qualifier and combined
the corpus / outcome claim into one sentence in the FAQ. Per GPT:
the corpus is INTERNAL benchmark scenarios (not real market data),
and the score / outcome should be a separate sentence so the
reader can see the framing clearly.
Fixed:
- trust card: "Validated against six adversarial conveyancing
scenarios" -> "Validated against six internal adversarial
conveyancing benchmark scenarios".
- FAQ wrapper: split the corpus claim and the outcome claim into
two sentences. "The current corpus of six internal benchmark
scenarios averages 99 out of 100; in that corpus, every planted
issue was detected and there were zero over-flags of routine
items."
Both numbers (6, 99) remain grep-verifiable in BASELINES.md.
Refs: GPT external review of v3 bench, "marketing wording: slightly
tighten it" + "one thing to fix immediately";
feedback_marketing_must_reflect_truth.md.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
GPT review caught two truth issues in the v3 results. Both fixed.
Truth issue 1: wrong uncovered-codes count
Previous wording in BASELINES.md "honest gaps" + PR #17 description: "5 of 21 taxonomy codes unexercised". Actual count, verified by grepping
EXPECTED_ISSUE_CODESagainst the fullConveyancingIssueCodeunion: 13 exercised, 8 uncovered.The list of 8 uncovered:
Per GPT: "Especially important:
TITLE_DEFECTIVE_INDEMNITY_REQUIRED,SEARCH_CONTAMINATED_LAND,MORTGAGE_OFFER_EXPIRY_RISK,LEASE_ALTERATIONS_CONSENT- these sound commercially valuable because they map to real lender/client concern." Captured as the priority order for any future speculative growth.Truth issue 2: marketing copy missing "internal" qualifier + single sentence too dense
GPT's tightening suggestion: signal that the corpus is internal (not real market data), and separate the corpus claim from the outcome claim.
Trust card ("Severity calibrated"):
FAQ wrapper answer:
Both numbers (6, 99) remain grep-verifiable in BASELINES.md.
What this changes
packages/fixtures/BASELINES.md: new "Honest gaps in the corpus" section listing the 3 unrebenched legacy scenarios and the 8 uncovered codes, with speculative-priority order for future growth.apps/marketing/src/sections/trust.tsx: "Severity calibrated" card body gets "internal" + "benchmark" qualifiers.apps/marketing/src/sections/faq.tsx: wrapper answer splits the corpus claim and the outcome claim into two sentences.Test plan
pnpm typecheck- 9/9 workspaces passpnpm check- 0 errors, 3 pre-existing warningspnpm build- 3/3 tasks passCONVEYANCING_ISSUE_CODESandEXPECTED_ISSUE_CODESBrand and liability check
Moat level
L7 - trust signal. Investors / partners / COLPs reading the trust page get a more accurate framing of where the validation stands.