Performance work on a trading system has an unusual rule. A faster system that decides differently is a different system, and all the research behind it no longer applies. So the goal in October 2026 was narrow: reduce the time the platform spends on each cycle and prove that nothing else changed.
NOVAQ's median internal execution cycle stood at 16.5 seconds. Six batches of work took it to 8.4 and then 2.77 seconds, in the windows outside heavy provider work. One frequent query fell from about 2.9 seconds to a few milliseconds, and session analytics from about 115 to 23 seconds per instrument.
How identity was protected
Each batch came with A/B identity checks on outputs and tests that had to fail before the change and pass after it. Measurement windows were fixed in advance. The later batches carried three signatures. The work ran through the AI audit swarm, which separated the role that made the change from the role that verified it.
Median internal cycle, October 2026 (seconds)
Why it matters
Short holding periods and tight signals depend on time. A cycle under three seconds changes what is practical for the system.
Limits
The 2.77 seconds applies outside the windows with heavy provider work, and later-session spikes remained. This is the platform's internal cycle, not order execution at the broker, and it makes no claim about returns.