card, license and attribution for release 2026-08-02

CITATION.cff

@@ -7,11 +7,10 @@
 # in DATASET_VERSION (dataset/schema.py) and is recorded inside each release
 # manifest; a citation needs to name the bytes, and only the release id does.
 #
-# Release 2026-07-30 is built and verified (35,097,678 rows across
-# fund_holdings and insider_transactions, plus coverage; 63 partitions;
-# 873,967,532 bytes; verify_release CLEAN). It has NOT been uploaded anywhere
-# yet, so there is no DOI and no `repository-artifact` URL. Add one when it is
-# published; a citation to a release nobody can fetch is not much of a citation.
+# Release 2026-08-02 is built and verified: 36,941,679 rows across six public
+# tables, 650 partitions, 1,040,640,412 bytes, verify_release CLEAN. It
+# supersedes 2026-08-01 and includes the corrected timezone interpretation for
+# unzoned SEC complete-submission acceptance timestamps.
 cff-version: 1.2.0
 message: "If you use this dataset, please cite it as below."
 type: dataset
@@ -33,8 +32,8 @@ authors:
     given-names: Jack
     email: jackwlutz1@gmail.com
     affiliation: Candella Labs
-version: "2026-07-30"
-date-released: "2026-07-30"
+version: "2026-08-02"
+date-released: "2026-08-02"
 # NOT CC0. This file declared `CC0-1.0` until 2026-07-30, which is precisely the
 # misstatement LICENSE-DATA exists to prevent: a CC0 dedication is an act by a
 # rights HOLDER, and under Feist v. Rural Telephone we hold no copyright in the

README.md

@@ -31,27 +31,27 @@ configs:
   default: true
   data_files:
   - split: train
-    path: releases/2026-08-01/congress_trades/*.parquet
+    path: releases/2026-08-02/congress_trades/*.parquet
 - config_name: house_filings
   data_files:
   - split: train
-    path: releases/2026-08-01/house_filings/*.parquet
+    path: releases/2026-08-02/house_filings/*.parquet
 - config_name: senate_filings
   data_files:
   - split: train
-    path: releases/2026-08-01/senate_filings/*.parquet
+    path: releases/2026-08-02/senate_filings/*.parquet
 - config_name: fund_holdings
   data_files:
   - split: train
-    path: releases/2026-08-01/fund_holdings/*.parquet
+    path: releases/2026-08-02/fund_holdings/*.parquet
 - config_name: insider_transactions
   data_files:
   - split: train
-    path: releases/2026-08-01/insider_transactions/*.parquet
+    path: releases/2026-08-02/insider_transactions/*.parquet
 - config_name: coverage
   data_files:
   - split: train
-    path: releases/2026-08-01/coverage/*.parquet
+    path: releases/2026-08-02/coverage/*.parquet
 ---
 
 # knowable
@@ -60,7 +60,7 @@ US government financial disclosures as a point-in-time dataset: congressional
 trades, SEC Form 4 insider transactions, and 13F fund holdings. Every row carries
 **the instant it actually became public** and a **measured coverage denominator**.
 
-**35,213,599 rows. 602 Parquet partitions. 958 MB zstd. Release `2026-08-01`.**
+**36,941,679 rows. 650 Parquet partitions. 1.041 GB zstd. Release `2026-08-02`.**
 
 Immutable: corrections ship as a new release naming this one in `supersedes`,
 never as a rewrite of these bytes.
@@ -80,7 +80,7 @@ import duckdb
 
 duckdb.sql("""
     SELECT filer_name, ticker, side, amount_range_low, published_at
-    FROM 'hf://datasets/jwlutz/knowable/releases/2026-08-01/congress_trades/*.parquet'
+    FROM 'hf://datasets/jwlutz/knowable/releases/2026-08-02/congress_trades/*.parquet'
     WHERE ticker = 'NVDA' AND published_at <= '2026-01-01'
     ORDER BY published_at DESC
 """)
@@ -136,7 +136,7 @@ resources: congress/trades, house/filings, senate/filings,
 | table | rows | one row is |
 |---|---|---|
 | `fund_holdings` | 34,030,485 | one 13F position line |
-| `insider_transactions` | 1,067,193 | one Form 4 transaction line, all codes |
+| `insider_transactions` | 2,795,273 | one Form 4 transaction line, all codes |
 | `congress_trades` | 72,140 | one disclosed trade (House 55,796, Senate 16,344) |
 | `house_filings` | 41,380 | one House filing with its publication clock |
 | `senate_filings` | 2,394 | one Senate eFD document with its clock |
@@ -220,7 +220,7 @@ Newest filing in this release published `2026-07-29`.
 
 ### Source errors are carried, not silently repaired
 
-63 of 72,140 rows (0.118%) carry a `transaction_date` **after** their own
+79 of 72,140 rows (0.110%) carry a `transaction_date` **after** their own
 publication instant, which is physically impossible. They are filer typos: years
 written as 3031, 2202, 2220. They ship as filed. Normalization reads the source's
 vocabulary; it does not repair the source. Filter `transaction_date <=
@@ -248,7 +248,7 @@ to waive.
   title  = {knowable: US government financial disclosures with a publication clock},
   author = {Lutz, Jack},
   year   = {2026},
-  note   = {Release 2026-08-01},
+  note   = {Release 2026-08-02},
   url    = {https://huggingface.co/datasets/jwlutz/knowable}
 }
 ```