TL;DR: Objection Academy's trial and jury scenario generation is now substantially faster. A complex production trial scenario that previously took about 107 seconds now completes in about 6 seconds, while preserving a deep, interactive decision tree.
Faster feedback, without flattening the exercise
Trial practice is most useful when a decision creates a consequence quickly enough to keep strategic momentum. Waiting more than a minute for a generated case can turn a focused practice session into a loading screen.
Objection Academy's case-generation backend has been rebuilt around that problem. The system now uses Gemini 3.5 Flash Lite through OpenRouter for fast structured generation, with a single Gemini 3.5 Flash fallback when a response needs extra recovery. The change replaces repeated attempts with a deliberate quality check, so the system spends time generating a useful tree instead of repeating an unsuccessful request.
What changed
The update makes three practical improvements.
- One focused attempt per model. The previous path could make several long requests before returning a result. The new path limits each model to one attempt.
- Depth is validated. A trial tree must meet the minimum depth requirement before it is accepted. If the fast model returns a shallow tree, the fallback gets one opportunity to produce a stronger result.
- The output is designed to fit. The generator now asks for a compact strategy spine with concise branch text. That leaves enough room for consequential choices, outcomes, and realistic trial tactics without truncating the JSON response.
The result in production
On a complex federal civil-rights scenario involving body-camera footage, spoliation, hearsay, expert testimony, impeachment, and qualified-immunity themes, the prior production path completed in about 107 seconds. The updated production path completed in about 6 seconds.
The faster response was not a smaller exercise. The live test returned 19 branches and a reachable path 10 levels deep. A simpler tenant-and-security-deposit scenario also returned 19 branches and a 10-level path in about 7 seconds.
Jury-selection generation improved as well. A complex 24-juror case completed in about 6 seconds with a valid case summary and juror records, while a focused juror voir dire scenario completed in just over 2 seconds with a three-choice opening decision.
Why depth still matters
A useful trial simulator cannot stop at a single question and answer. It needs to let a user pursue a theory, confront a setback, choose another tactical route, and see how the next decision changes the case.
The new generation contract intentionally creates a main strategic path through the scenario while preserving alternate outcomes. That structure supports more meaningful repetition: try one approach, replay the situation, and test the next decision without waiting for an entirely new case to arrive.
Built for the next decision
The goal is not to make a timer look better. It is to keep attorneys and law students in the decision-making loop. Faster generation means more opportunities to test a theory, refine an objection, reconsider a challenge, and keep practicing under realistic courtroom pressure.
Objection Academy will continue measuring both response time and scenario quality as the simulator evolves. The standard is simple: deep enough to be useful, fast enough to keep practicing.