Self-Updating Context
Context that stays current on its own
Shanc watches your warehouse schema, git history and query logs continuously. When something changes — a column renamed, a model rewritten, a metric redefined — the affected definitions update themselves. Your agents work from today’s business logic, every day.
- Continuous monitoring across schema, git and query patterns
- Changes scoped to exactly the definitions they affect
- Zero manual curation — no analyst maintaining the layer by hand
How it works
Continuous monitoring
Always watching
Shanc tracks the same sources it built your layer from. A migration lands, a dbt model changes, analysts start querying a table differently — each is an event the layer sees the moment it happens.
Drift detection
Shanc knows which definitions a change touches
A renamed column can sit behind a dozen metrics. Shanc maps each change to the definitions that depend on it, so an update is scoped and precise — the rest of the layer stays untouched and trusted.
Definitions update themselves
Current context, without the upkeep
Affected definitions are rebuilt and served immediately — agents get the new logic on their next question. The maintenance work that keeps analysts curating docs and semantic layers by hand runs in the background instead.
Shanc in benchmarks
95% accuracy on a live 118-table warehouse — and it holds as the warehouse changes.
We measured Shanc against agents working from raw schema on a real production warehouse. Agents with Shanc’s context layer passed 95% of tests at full coverage, against 73% for Claude Code at the same coverage. Accuracy like this comes from complete, current definitions — the state a self-updating layer is always in.
View the benchmarkContext goes stale faster than anyone maintains it
Fivetran’s 2026 enterprise benchmark found 53% of data engineering capacity goes to maintaining and troubleshooting existing pipelines rather than building new capabilities. Keeping things current by hand is already half the job. (Fivetran Enterprise Data Infrastructure Benchmark, 2026)
The average enterprise runs 300+ pipelines and sees 4.7 failures a month, each taking nearly 13 hours to resolve — with schema changes among the top causes. Drift is the background condition of a warehouse, and every unnoticed change quietly ages the context around it. (Fivetran · BusinessWire)
71% of data professionals cite incorrect or hallucinated outputs reaching stakeholders as a top concern, and trust in data jumped from 66% to 83% year over year as an organizational priority — the steepest rise of any measured objective. A stale context layer produces exactly this failure: confident answers built on last quarter’s definitions. (dbt Labs, State of Analytics Engineering 2026)
A context layer built once starts aging the day it ships. One that maintains itself compounds instead.