Measured accuracy
Every month we draw a random sample of published facts and check each one against its live source. We publish the result, including any errors we find.
November 2026
0 errors in 29 facts checked (error rate 0.0%; 95% confidence the true rate is between 0.0% and 11.7%). 1 more could not be confirmed from the cited source and are listed below.
Drawn from 687 sourced facts on published records.
| Month | Checked | Errors | Error rate | 95% range | Unconfirmed |
|---|---|---|---|---|---|
| November 2026 | 29 | 0 | 0.0% | 0.0% to 11.7% | 1 |
How it works
- What is sampled: every fact that carries a source link on a published record: prices, plan terms, what syncs, limits, requirements and agent install commands.
- How it is drawn: a random sample spread across fact types in proportion to how common they are. The draw is seeded from the month, so anyone with the data can reproduce it.
- How it is checked: each fact is compared with its cited source as it reads that day. A fact is an error if the source now says something different. If the source does not state the fact, we mark it unconfirmed and fix the record.
- What happens next: errors and unconfirmed facts are corrected, and the correction appears in the change history.
Raw results: /api/accuracy.json. Found a mistake? Email corrections@trylode.io.
Not confirmed by the cited source (November 2026)
- convex--vercel regions.supported ("GLOBAL"): The cited marketplace page does not mention regions. GLOBAL was recorded from the absence of any restriction; needs an explicit source. Fixed 14 Nov 2026: every Vercel record now shows regions as 'No regional restriction stated' (regions.basis = no-restriction-stated).