Skip to content

docs: the benchmark corpus holds 15,926 events, not 16,146 - #34

Merged
zz-plant merged 1 commit into
mainfrom
claude/fix-corpus-event-count
Oct 8, 2026
Merged

zz-plant merged 1 commit into
mainfrom
claude/fix-corpus-event-count

Conversation

@zz-plant

@zz-plant zz-plant commented Oct 8, 2026

Copy link
Copy Markdown
Collaborator

Summary

README, BENCHMARK.md and the CHANGELOG said the benchmark-corpus-2026-08-27 release holds 16,146 events. The manifest committed with it shows 15,926:

  • scripts/build-benchmark-corpus.sh writes each page's events from wc -l of its event file, one event per line.
  • The manifest's SHA-256 hashes bind the release asset.
  • Its ten counts sum to 15,926: Donald Trump 8,317, COVID-19 pandemic 3,828, Russia 1,395, Earth 599, CRISPR 447, Climate change 409, Bitcoin 354, Tesla, Inc. 353, COVID-19 202, GPT-4 22.

The 16,146 was typed by hand into the publishing commit (b5201fe) and its docs, 220 high. This PR changes the three prose figures only; the data, manifest and release are untouched.

Related issue

N/A

Type of change

  • Deterministic (analyzer, ingestion, event types)
  • Ground truth / eval
  • CLI / tooling
  • Documentation
  • Infrastructure / CI

Verification

Docs-only change; no code, build input or test fixture is touched. The manifest sum was recomputed from benchmark/ground-truth/manifest.json. The one test that reads README.md (packages/ingestion/src/__tests__/sources.test.ts) uses it as a sample file for git revision history, not for its content.

  • bun run build passes (not run: docs only)
  • bun run typecheck passes (not run: docs only)
  • bun run test passes (not run: docs only)
  • bun x biome ci packages/ passes (not run: no files under packages/ changed)
  • New analyzers include an eval (n/a)
  • Architecture changes update ARCHITECTURE.md (n/a)

Scope check

  • I read docs/repository-boundary.md
  • Does not target individual editors, score sentiment, predict outcomes, automate editing, or make truth claims
  • Does not add healthcare-specific logic, domain-specific source weighting, or customer-workflow rules

🤖 Generated with Claude Code

https://claude.ai/code/session_01CHTDKD1277KYSjDF1gpnLD


Generated by Claude Code

README, BENCHMARK.md and the CHANGELOG all said the
benchmark-corpus-2026-08-27 release holds 16,146 events. The manifest
committed alongside it says otherwise: scripts/build-benchmark-corpus.sh
writes each page's count from `wc -l` of its event file (one event per
line), the manifest's hashes bind the release asset, and its ten counts
sum to 15,926. The 16,146 was typed into the publishing commit's
message and docs by hand, 220 high. The data and manifest are
unchanged; only the prose was wrong.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01CHTDKD1277KYSjDF1gpnLD
@zz-plant
zz-plant merged commit 538caf8 into main Oct 8, 2026
3 checks passed
@zz-plant
zz-plant deleted the claude/fix-corpus-event-count branch October 8, 2026 16:20
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants