The database that never reads your data
Most of the columns in a real table aren’t data — they’re a function: autoincrement ids, evenly-spaced timestamps, accumulating counters. AT-1 recovers the function and answers a range predicate and its exact selectivity from a few coefficients, in microseconds, touching zero stored bytes.It isn’t a faster scan — it’s a different physics of query.
- 30,361×
- autoincrement id, selective range vs full scan
- 61,770×
- evenly-spaced timestamp predicate
- 0 bytes
- stored values read on the pure analytic path
- = DuckDB
- byte-identical results, DuckDB as oracle (19/19)
What a scan costs vs what GenQuery costs
-- SELECT COUNT(*) FROM events WHERE id BETWEEN 4000000 AND 4000100 (5M-row table) DuckDB / scan: 904.75 ms reads the column AT-1 GenQuery: 29.8 us reads 0 bytes -> 30,361x, byte-identical answer -- and the EXACT selectivity is free, up front (no histogram, no sample): events.id BETWEEN a AND b -> matched=101, selectivity=0.002% (from the closed form)
The optimizer plans joins on truth, not an estimate— exact selectivity with zero histograms. Scrambled generators (an LCG/LFSR whose range scatters) fall back to a normal scan: correct, just not accelerated. Every result is validated byte-identical against DuckDB (19/19); nothing here is approximate.
ids, timestamps, counters, sequences — the dominant columns in logs, events, telemetry, and ledgers are all generator-produced.
the query planner gets the true matching-row count from a few coefficients — no histogram, no sample, no scan.
it's built into the AT-1 database engine and the Postgres wire — your existing SQL and BI tools get it for free.