feat: the heatmap says what management was actually naming

feat: Ask AI opens with the question an analyst would actually ask

feat: the filed statements, and tone measured instead of quoted

feat: The Brief renders as the published research note

fix: close the numpy import race that took the live Space down

feat: the guidance record fills the half of Track Record that never worked

feat: score management's own revenue ranges against what it then filed

docs: three guidance-extraction defects, found by reading the data back

docs: the Space is now Viney/readthrough

docs: how to publish to each remote, and a script for the awkward one

data: GOOGL Q12023 recovered, 88 of 89 rows now verified

docs: close both defects in the run record, and correct one classification

fix: the two coverage defects the six-ticker backfill exposed

data: AMD brief generated locally on Opus 5 against the repaired tables

data: refreshed MD&A and risk-factor sections from the re-extraction

data: re-extract 88 periods, guidance from the calls, segments on six tickers

fix: re-extract the stored period set, not EDGAR's newest window

test: pin the enumerate-don't-fail invariant for new statement lines

feat: the acceptance criteria as a runnable report

test: close the gap in the no-Alpha-Vantage guarantee

feat: re-extraction mode that repairs rows without touching Chroma

fix: prefer the total concept over its narrow slice for two lines

feat: read eight statement lines already in the downloaded payload

fix: the gross-profit gate must match the filing, not just the period

fix: the two fallback_non_sec causes, one derivation and one concept switch

fix: a guidance period with no figures is silence, not a parse

feat: guidance carries its source quote and a distinguishable outcome

test: make the budget-independence test detect a shared budget

test: make the guidance regression test actually reproduce the bug

fix: guidance parser read the operator's introduction, not the outlook

docs: implementation plan for the data foundation (sub-project 1)

docs: correct Workstream C — the two fallback causes, diagnosed

docs: analysis-depth roadmap and the data-foundation spec behind it

docs: design spec for falsification conditions (sub-project D)

feat: decompose the EPS headline and read the statements behind it

data: segment tables, the Readthrough brief and the run cost log

chore: rename the project to Readthrough

feat: rebuild The Brief around a sourced article and scannable panels

feat: read per-product and per-region results from the filing's XBRL instance

feat: add Claude Opus 5 to the model catalogue

Merge branch 'sdd/prospection': buy-side prospection sub-system

docs: preserve the prospection execution decision log

fix(prospection): close the three blocking defects and seven Important ones from the whole-branch review

fix(prospection): pass 1 gates on declared equity, not a confirmed-holding overlap

docs: fix the real bootstrap block - pass 1 gated on a value it cannot have

fix(prospection): break the Tier 1 bootstrap cycle, correct CLAUDE.md coupling claim

docs: rule on the Task 14 findings - the Tier 1 bootstrap circularity

docs: separate enforced refusals from human process rules in the prospection README

docs(prospection): playbook, GDPR processing record, end-to-end suite

fix(prospection): close five CLI/eml gaps found in review