Skilirium

Recognises a business date stored with a time of day

K-828 · Skill knobit · Activity: Generate and review DDL · Competence: Data modelling

Two mirror-image defects. An event timestamp with no offset is ambiguous twice a year and incomparable across a bank that books in more than one zone. A business date — value date, booking date, maturity date, due date — stored as a timestamp invites a time component nobody means and makes date equality quietly fail.

Can AI do this?

AI does this partly Some of the tested AI models master this; others do not show it. 3 of 10 models master it.

ModelTierProbability of masteryStatus
gemini-4-argonFlagship 98%Mastered
gemini-3.1-pro-previewFlagship 14%Not shown
gpt-6-astraFlagship 20%Not shown
claude-fable-5-1Flagship 98%Mastered
gemini-3.8-flashMid tier 14%Not shown
gpt-6.1-solMid tier 18%Not shown
claude-sonnet-5-5Mid tier 99%Mastered
gemini-3.5-flash-liteLow cost 15%Not shown
gpt-6-lunaLow cost 19%Not shown
claude-haiku-5-5Low cost 14%Not shown

Mastered means a probability of at least 90% that the model has this knobit, estimated from its graded answers. “Not shown” means the answers do not settle it either way. How the models are tested.

What it looks like in practice

Expected result: The column and the date type it needs

How it is assessed: Flags the deficient case and passes the acceptable one. Deficient: created_at TIMESTAMP with no zone on dim_account, written by jobs in two regions. And payment_value_date as a TIMESTAMP, so = DATE '2026-03-31' matches none of the rows loaded at 23:30 Acceptable: TIMESTAMP WITH TIME ZONE for event times, or UTC with the convention stated in the column comment and nowhere else; DATE for payment_value_date, loan_maturity_date and valid_from_date/valid_to_date, as the sample model has them

Also known as: Point in time without a zone, business date with a clock (generated DDL)

Related knobits

Builds on

All 33 knobits of Generate and review DDL.

Source: Review checklist 08-ddl, row 18; ISO 8601; ISO 9075