for seed in 1 2 3 4 5 6; do RUN_DIR="results/progressive_C11_target24_seed${seed}_$(date -u +%Y%m%dT%H%M%SZ)" mkdir -p "$RUN_DIR"

python3 -u scripts/search_progressive.py
--group C11
--target-distance 14
--seed "$seed"
--classical-distance-backend fast
--quantum-steps-fast 3000
--quantum-steps-slow 300000
2>&1 | tee "$RUN_DIR/run.log" done

By default, scripts/search_progressive.py now updates best_codes/ at the end of a successful run (including publishing website data and pushing to GitHub). Use --no-best-codes-update to disable this, or pass --no-git / --no-publish to limit the post-run steps. If you stop a run with Ctrl-C, it will still attempt the best-codes update.

Quick smoke integration test (skips git/publish + history scan):

SMOKE_BEST_CODES_UPDATE=1 bash scripts/smoke_progressive_search.sh

Update best_codes (scrape + publish)

Dry run (no filesystem or git changes):

python3 scripts/scrape_and_publish_best_codes.py --dry-run

Full update (sync best_codes/, rebuild data.json/index.tsv, commit + push):

python3 scripts/scrape_and_publish_best_codes.py

Common flags:

• --no-git skip commit/push
• --no-publish skip website data updates
• --verbose show skipped files + actions

Note: the scraper now scans git history to recover older best codes, so a run can take about a minute.

Matrix download tracking (GitHub Releases)

• Website matrix links point to GitHub release assets tagged per track: best-codes-matrices-633, best-codes-matrices-844, best-codes-matrices-212.
• Publish (or sync) release assets:

python3 scripts/publish_matrices_release.py --dry-run
python3 scripts/publish_matrices_release.py

• Report download counts:

python3 scripts/release_download_counts.py

Workflow

See docs/PROGRESS_LOG.md for the current project state, canonical scripts, and the end-to-end workflow (search -> scrape/publish -> refine).

Two-machine setup

• Keep a dedicated maintenance clone for publishing; do not git pull into a running search clone.
• Archive untracked run artifacts into local_results/ before syncing.
• Run website/best_codes publish steps from the maintenance clone only.

Refine then publish

• Refinement scripts (scripts/refine_best_codes.py and the length/group pipelines) update metadata and then resync/publish best_codes by default.
• If you run custom refinement steps, follow with scripts/scrape_and_publish_best_codes.py to rebuild best_codes/data.json and best_codes/index.tsv.

Local code tracks

• Default track uses local [6,3,3] codes and publishes to best_codes/.
• The [8,4,4] track uses --local-a 8_4_4 --local-b 8_4_4 and publishes to best_codes_844/ (with pending artifacts under codes/pending_844/).

Context for new contributors / new chats

• docs/CHAT_CONTEXT.md
• docs/CHAT_PROMPT_TEMPLATE.md
• docs/ARCHITECTURE.md