Skip to content
CN2.DEVBUKYA NARESH

LAB / REPLAY-01

OPERATIONAL

POINT-IN-TIME REPLAY LAB

The signature demonstration: select a timestamp on a synthetic series and observe the information boundary — what was knowable at t, and what only existed after.

IN — TIMESTAMP · SCRUB · REPLAY SPEED

OUT — VISIBLE AT t · FUTURE-DATED · LOOKAHEAD DEMO

INSTRUMENT / REPLAY-01

INPUTS DECLARED · OUTPUTS DERIVED

Point-in-time replay — information boundary at t = 23
VISIBLE AT t: 24 CANDLESFUTURE-DATED: 18 CANDLES

A query at t can only read the solid candles. The ghosted region is information that did not yet exist — in a naive backtest it silently leaks in (lookahead); in point-in-time infrastructure it is structurally unreachable. Series is deterministic and synthetic.

LOOKAHEAD VS POINT-IN-TIME SAFE

THE FAILURE MODE

LOOKAHEAD

  • A resampled candle quietly includes the tail of its window
  • A join keys on knowledge that only existed after the fact
  • A timestamp normalizes to a coarser clock and rewrites causality
  • The result cannot be repaired — only regenerated from clean data

THE STRUCTURAL ANSWER

POINT-IN-TIME SAFE

  • Reads at time t can only resolve information that existed at t
  • The guarantee lives in the data layer — consumers cannot opt out
  • Sliding watermark defines the visible horizon for every reader
  • Adversarial tests attack the boundary and fail — verified
SEE THE IMPLEMENTATION →