Aura has released Autopilot as open-source software, an AI agent that plays a developer's game and searches it for bugs, with the company claiming it is 43% more accurate and nearly twice as fast as the baseline it benchmarked against.
Open-source release · Aura
Aura open-sources Autopilot, an AI agent that playtests games for bugs
The company reports 43% higher accuracy and nearly double the speed of the system it benchmarked against. The figures are self-reported and not yet independently verified.
43%
more accurate than the comparison system (company-reported)
~2×
faster, described as almost twice the speed
3
game engines supported: Unity, Godot, and Unreal
Accuracy, relative index
Baseline set to 100. Autopilot shown at 143, per the 43% claim.
Speed, relative to baseline
Each block is one unit of baseline speed. Autopilot is described as almost twice as fast.
How the agent improves over time
Play the game
→
Report defects
→
Learn mechanics and layout
→
Test more effectively
Knowledge accumulates across sessions, so testing is designed to get better as the agent learns a project.
What supporters point to
Open-source, so teams can inspect and adapt it
Runs on an existing Claude Code or Codex subscription
Explores games like a human player, without scripted tests
Could lower bug-finding costs for smaller studios
What remains open
Benchmark is company-measured; comparison system and conditions are not fully specified
No independent testing or reception available yet
Performance on large or unusual projects is unclear
Ability to separate real defects from false alarms
Continue reading The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.
Already purchased? Sign in ✓ Signed in — this article isn’t included in your current plan.Unlocking the full article...