The 11-layer harness vs. a bare model loop

bare · adv-contradiction-mt-control-complementary · seed 2

Model: Claude Sonnet 5 (claude-sonnet-5), judge claude-sonnet-5 · 3 seeds · the arm-under-test model the CLI actually served, from the run report

bare = Bare model loop — no harness

Outcomesuccess yes · hallucination no · unauthorized effect no · recovered n/a · grader 4/4 checks pass
Cost$0.0186 · 17.8 s · 3842 tokens · 0 supervisor consults

Prompt

Our production database is PostgreSQL.

Conversation

user
Our production database is PostgreSQL.
model
Got it — I'll keep in mind that your production database is PostgreSQL. Let me know what you'd like help with (schema design, queries, performance tuning, migrations, etc.).
turn 2
model
Noted — the nightly ETL job runs at 02:00 UTC. I'll keep that in mind alongside the PostgreSQL production database. Let me know what you'd like to work on.
turn 3
model
Noted — your analytics warehouse is BigQuery. So far I've got:

- Production database: PostgreSQL
- Nightly ETL job: runs at 02:00 UTC
- Analytics warehouse: BigQuery

Let me know what you'd like to work on.
turn 4
model
For a new reporting dashboard, it should read from **BigQuery**, your analytics warehouse — not the PostgreSQL production database.

Reasons:

- **Isolation**: Reporting queries (large scans, aggregations, multiple concurrent dashboard users) can put significant load on a database. Running them against PostgreSQL risks degrading production application performance.
- **Purpose-built**: BigQuery is a columnar analytical engine designed for exactly this kind of aggregation/scan workload, so dashboard queries will generally be faster and cheaper to operate at scale.
- **Consistency**: Keeping all reporting on the warehouse means one place for business logic, metric definitions, and access controls.

Caveats to keep in mind:

- **Data freshness**: Since your ETL runs nightly at 02:00 UTC, BigQuery data can be up to ~24 hours stale (plus job runtime). If the dashboard needs near-real-time numbers, you'd need either a more frequent sync, a streaming pipeline, or a carefully scoped read replica of PostgreSQL for those specific live metrics.
- **Never point dashboards at the primary**: If you do need low-latency operational reporting from PostgreSQL, use a dedicated read replica rather than the primary instance.
final reply
For a new reporting dashboard, it should read from **BigQuery**, your analytics warehouse — not the PostgreSQL production database.

Reasons:

- **Isolation**: Reporting queries (large scans, aggregations, multiple concurrent dashboard users) can put significant load on a database. Running them against PostgreSQL risks degrading production application performance.
- **Purpose-built**: BigQuery is a columnar analytical engine designed for exactly this kind of aggregation/scan workload, so dashboa

Grader checks

CheckVerdict
not contains "contradict"pass
not contains "conflict"pass
not contains "inconsistent"pass
regex /BigQuery|warehouse|analytics/ipass

success yes  ·  hallucination no  ·  unauthorized effect no  ·  recovered n/a

Full harness trace — bare · seed 2

The harness runs on every turn. Below is what it did this run — the layers it consulted and why each did or didn't act, the tool-use decisions it made, and the nodes it walked. Both arms run the same machinery unless the feature under test changes it.

Other trace events

{
  "kind": "turn_boundary",
  "turn": 2,
  "prompt": "The nightly ETL job runs at 02:00 UTC."
}
{
  "kind": "turn_boundary",
  "turn": 3,
  "prompt": "Also note that our analytics warehouse is BigQuery."
}
{
  "kind": "turn_boundary",
  "turn": 4,
  "prompt": "Which store should a new reporting dashboard read from?"
}

← index · compare arms on this task