v0.1 · Now shipping benchmarks

What changed

Release notes, new features, and fixes. Shipped in order, without the noise.

Peer benchmarks — see where you stand

Anonymous cohort benchmarks are now live for all connected organizations. Compare your AI tool spend, license utilization, and bug-density ratios against companies at the same headcount and stage — no data leaves your environment unblinded.

Benchmark cards appear on the Financial Operations and Code Health dashboards. Each card shows your value against the P25, P50, and P75 marks for your cohort. Cohort membership is computed weekly; you will see your first benchmarks within 7 days of connecting your first data source.

GitHub App and Copilot Metrics integration

Connect your GitHub organization in under 5 minutes. Backlight ingests pull request metadata, review cycles, and merge rates via the GitHub App — no PAT rotation, no manual exports.

GitHub Copilot Metrics API is now supported. Seat-level utilization, active-user counts, and acceptance rates sync automatically every 24 hours. Idle seat detection fires on any seat that has not accepted a suggestion in 14 days.

Bug: fixed an off-by-one error in the sprint boundary calculation that caused the last day of a sprint to be attributed to the following sprint in bug-density charts.

Backlight v0.1 — first light

The first public release of Backlight. Built for engineering leaders who need an honest read on their AI coding tool spend — not vendor marketing, not anecdote, not a spreadsheet maintained by hand.

Ships with:

  • AI attribution engine — classifies every merged commit as AI-coauthored, AI-assisted, or human-only using heuristics against commit metadata
  • Financial Operations dashboard — spend, license utilization, idle seat detection, and cost-per-PR by vendor
  • Code Health dashboard — rewrite rate, bug density by attribution, and regression clustering across the last 8 sprints
  • Alerting — configurable thresholds on rewrite rate, idle seats, and spend variance