Community re-run · simulated
“Does this prompt still hold?”
Models move. A published fidelity average is a snapshot from the day it ran, so the interesting question is always whether a brief still reproduces. Pick a prompt and a model below: the studio streams a run log and compares its simulated score with the published average.
This page runs a simulation — no model is called and no tokens are spent. Its purpose is to show the shape of the feature: the run-log format, the comparison against published numbers, and the honesty about which log is real. Real runs are the ones recorded on each prompt page.
You came here from furniture-maker-joinery, so that prompt leads the list below.
Community re-run · Furniture maker — joinery and honesty
Anyone can ask “does this still hold?” — pick a model and watch the run log stream, then compare with the published average.
Run log
// press Re-run to stream the simulated logComparison
The queue keeps your runs in memory only, and the log is generated locally — the fidelity figures are a deterministic simulation, not a measurement. Published run logs on each prompt page remain the real record.
Community re-run · AI tool landing with terminal motif
Anyone can ask “does this still hold?” — pick a model and watch the run log stream, then compare with the published average.
Run log
// press Re-run to stream the simulated logComparison
The queue keeps your runs in memory only, and the log is generated locally — the fidelity figures are a deterministic simulation, not a measurement. Published run logs on each prompt page remain the real record.
Community re-run · A11y-strict page — WCAG AA as a hard requirement
Anyone can ask “does this still hold?” — pick a model and watch the run log stream, then compare with the published average.
Run log
// press Re-run to stream the simulated logComparison
The queue keeps your runs in memory only, and the log is generated locally — the fidelity figures are a deterministic simulation, not a measurement. Published run logs on each prompt page remain the real record.
Community re-run · Dark SaaS launch page with aurora hero
Anyone can ask “does this still hold?” — pick a model and watch the run log stream, then compare with the published average.
Run log
// press Re-run to stream the simulated logComparison
The queue keeps your runs in memory only, and the log is generated locally — the fidelity figures are a deterministic simulation, not a measurement. Published run logs on each prompt page remain the real record.
Community re-run · Rewilding campaign page
Anyone can ask “does this still hold?” — pick a model and watch the run log stream, then compare with the published average.
Run log
// press Re-run to stream the simulated logComparison
The queue keeps your runs in memory only, and the log is generated locally — the fidelity figures are a deterministic simulation, not a measurement. Published run logs on each prompt page remain the real record.
Why let anyone re-run
Prompt quality claims should be cheap to challenge. If a model update breaks a brief, the fastest way to find out is to let the people using it say so.
What the queue would do
A real implementation would rate-limit runs, pin model versions, and append the result to the prompt's run log with a community tag.
What happens today
Nothing is persisted. Your runs live in this tab's memory; refreshing clears them, and the published logs stay untouched.