Historical model audit

Forecast Replay

Return to a historical cutoff and reconstruct what 1stSeen could have predicted, using only the evidence it held by then. Every run is computed by the worker’s leak-safe backtest when you ask for it.

Future evidence is sealed

The held-out opening and every post-cutoff observation are scoring data only. They cannot enter the forecasting component.

What can be replayed

Held-out targets with a defensible actual interval

Roles with two or more recorded openings: 179
Excluded: only observed-by evidence: 26
Replay candidates: 153
held out on an exact opening: 153
held out on a bounded opening: 0

An archive capture proves a role was visible by then, not that it opened then, so an observed-by opening is never scored as ground truth.

Latest backtest

Why targets could not be scored

Finished Oct 3, 2026 · cutoff 60 days before each target · 890 targets, 0 scored, 890 skipped

no temporal evidence was available by the forecast cutoff689
held-out opening is observed-by only and has no scoreable actual interval201

Counted over every reconstructed cycle the backtest evaluated, which is more than the replay candidates above. Reasons are the backtest’s own words.

Choose a held-out target

Showing 141–153 of 153

Palantir
Software Engineer New Grad Defense
Held outJun 19, 2025exactexact
Skipped in the latest backtest
no temporal evidence was available by the forecast cutoff
Palantir
Software Engineer New Grad Defense
Held outJun 19, 2025exactexact
Skipped in the latest backtest
no temporal evidence was available by the forecast cutoff
Verkada
Software Engineer Platform Infrastructure
Held outJun 9, 2025exactexact
Not in the latest backtest run
Palantir
Product Designer New Grad
Held outNov 21, 2024exactexact
Skipped in the latest backtest
no temporal evidence was available by the forecast cutoff
Palantir
Software Engineer New Grad
Held outSep 24, 2024exactexact
Skipped in the latest backtest
no temporal evidence was available by the forecast cutoff
Palantir
Software Engineer Intern Defense Tech
Held outAug 5, 2024exactexact
Skipped in the latest backtest
no temporal evidence was available by the forecast cutoff
Palantir
Software Engineer Intern Defense Tech
Held outAug 5, 2024exactexact
Skipped in the latest backtest
no temporal evidence was available by the forecast cutoff
Palantir
Software Engineer Intern Defense Tech
Held outAug 5, 2024exactexact
Skipped in the latest backtest
no temporal evidence was available by the forecast cutoff
Palantir
Information Security Engineer Intern
Held outAug 3, 2022exactexact
Skipped in the latest backtest
no temporal evidence was available by the forecast cutoff
Palantir
Forward Deployed Software Engineer New Grad US Government
Held outJun 30, 2022exactexact
Skipped in the latest backtest
no temporal evidence was available by the forecast cutoff
Palantir
Software Engineer New Grad
Held outJul 1, 2021exactexact
Latest backtest held out another posting of this role that year
Palantir
Software Engineer New Grad
Held outOct 27, 2020exactexact
Skipped in the latest backtest
no temporal evidence was available by the forecast cutoff
Palantir
Privacy and Civil Liberties Software Engineer Intern
Held outOct 6, 2016exactexact
Skipped in the latest backtest
no temporal evidence was available by the forecast cutoff

Choose the historical vantage point

Palantir · Software Engineer New Grad Defense · held out Jun 19, 2025 exact

Forecast cutoff

Why targets could not be scored

Finished Oct 3, 2026 · cutoff 60 days before each target · 890 targets, 0 scored, 890 skipped

no temporal evidence was available by the forecast cutoff689

held-out opening is observed-by only and has no scoreable actual interval201

Counted over every reconstructed cycle the backtest evaluated, which is more than the replay candidates above. Reasons are the backtest’s own words.