This is a point-in-time record of the week of 2026-05-18 to 2026-05-24. It was written from the repository's history for that week and is not revised afterwards, so later pages may supersede what it says.

Week of 2026-05-18 to 2026-05-24

Window: 2026-05-18 00:00:00 to 2026-05-24 23:59:59, local author time Recapped: 2026-08-28 Commits in window: 10 (7 non-merge, 3 merges)

This is a backfill written 96 days after the window closed. All ten commits were authored on May 19 or May 20. The maintainer recalled on 2026-08-28 that this period was probably a vacation with some bug-fixing performed from Scotland. The recorded timestamps are consistent with that recollection: the May 19 work ran from 15:34 to 16:25 BST and the May 20 commits were at 18:27 BST when converted to Scotland time. Git retained a -05:00 author offset, however, so the times support the account without independently establishing the author's location.

A scan of all six public GitHub repositories on 2026-08-28 found no work in the other five during this week; all visible commits were in new_car_tracker.

What shipped

The week was a short DuckDB and MinIO troubleshooting burst. 52255f2 selected DuckDB's httplib HTTPFS implementation to revert behavior changed in DuckDB 1.5. Five subsequent commits tested the suspected parallelism and connection behavior by enabling HTTPFS connection caching, reducing dbt's DuckDB threads from four to two and then one, removing the explicit httplib selection, and finally restoring four threads (3587754, 3877444, 4864ee1, 5038c4b, f819bf0). The sequence records an experiment; its commit text does not state whether the underlying theory was confirmed.

Before pausing development, 202f448 added a 312-line LinkedIn case study and an SVG cover image. The same commit changed the then-current Plan 105 document marker from planned to complete, but its subject and body do not name a plan, so the commit remains unattributed under the recap's attribution rule.

What moved between states

Nothing moved in the planning index. docs/PLANS.md has no revision in the window. The status-marker edit inside the Plan 105 document at 202f448 is a document change, not an index state transition.

What is still owed

The DuckDB experiment ended with connection caching enabled, four threads restored and the explicit httplib setting removed. No commit in the window states whether the connectivity or parallelism issue was resolved, so the outcome cannot be recovered from this week's history alone.

As of 2026-08-28, Plan 105 is already present in the completed-plan archive; the historical marker change therefore needs no planning-state action today.

Unattributed commits

None of the seven non-merge commits names a plan in its subject or body:

• 52255f2 — revert to pre-1.5.x DuckDB behavior
• 3587754 — testing theory the issue is around parrallezation.
• 3877444 — testing theory the issue is around parrallezation.
• 4864ee1 — testing theory the issue is around parrallezation.
• 5038c4b — testing theory the issue is around parrallezation.
• f819bf0 — testing theory the issue is around parrallezation.
• 202f448 — adding docs before pausing development.

Merges

• 2f48ac6 — PR #127 from fix/duckdb_versioning; the branch names no plan.
• 1229679 — PR #128 from fix/duckdb_versioning; the branch names no plan.
• 52e3b1d — merge of master from the GitHub remote; no PR or plan is named.

Deferred to the next recap

641 commits sit after this window's end as of 2026-08-28.

git rev-list --count --all --since='2026-05-24 23:59:59'   # 641

That raw tail includes every following week through the current partial week. The complete weeks ending May 31, June 7, June 14, June 21, June 28, July 5, July 12, July 19, July 26, August 2, August 9, August 16 and August 23 are already covered. No following complete week is genuinely uncovered; the week ending August 30 is still in progress and is not yet recappable.