I Built an AI Mock Trial Platform Because Practicing Law Shouldn't Require a Full Cast

Jesper Deng — Wed, 20 May 2026 05:20:41 +0000

The Problem No One Talks About

If you're a law student preparing for mock trial, or a lawyer rehearsing for court, you face a frustrating reality: you can't practice alone.

A real trial involves a judge, opposing counsel, witnesses, and jurors. To run even a basic practice session, you need to coordinate 3-5 people's schedules. Most of the time, that just doesn't happen.

So what do people actually do?

Law students rehearse opening statements in front of a mirror
Mock trial teams only get 1-2 full run-throughs before competition
Young lawyers go into their first trial with almost no live courtroom experience
Solo practitioners have zero way to simulate cross-examination or hostile witnesses

The fundamental bottleneck isn't skill — it's access to practice partners.

What I Decided to Build

I asked a simple question: what if AI could play every other role in the courtroom?

Not a chatbot that answers legal questions. Not a document tool. A full courtroom simulation where:

You play the attorney (plaintiff or defense)
AI plays the judge — ruling on objections, managing procedure
AI plays opposing counsel — making arguments against you, objecting to your questions
AI plays witnesses — responding to direct and cross-examination with realistic personalities
The trial follows real procedure: voir dire → opening → witness examination → closing → verdict

At the end, you get scored on your performance.

Why This Is Harder Than It Sounds

The challenge isn't just "make AI talk like a lawyer." It's:

1. Multi-role coherence

The judge, opposing counsel, and witness are all AI — but they need to behave as separate people with different goals. The judge is neutral. Opposing counsel is adversarial. The witness has a backstory and may be unreliable. One model, multiple conflicting personas, in the same conversation.

2. Stage management

A trial isn't a free-form chat. It has strict procedural stages. You can't cross-examine during opening statements. The AI needs to enforce courtroom rules while still feeling natural.

3. Reactive complexity

If you raise an objection, the judge must rule. If sustained, opposing counsel must rephrase. If you introduce surprise evidence, the witness must react consistently with their backstory. Every action cascades.

4. Solo practice must feel real

If it feels like talking to a chatbot, lawyers won't use it. The responses need enough unpredictability and pushback to create genuine practice pressure.

What I Learned Building This

Start with the state machine, not the prompts. I spent too long tweaking AI personalities before realizing the real problem was managing trial flow. Once I built a proper stage system (8 stages, with rules for what's allowed in each), the AI behavior fell into place.

Scoring creates motivation. Early testers would quit mid-trial. Adding a verdict with performance scoring changed everything — people now complete full trials because they want to see their score.

Credits > subscriptions for this audience. Law students are broke. A generous free tier with credits lets them actually use the tool. Power users (practicing attorneys) will pay when they see the value.

Try It

If you're curious:

Live platform: mocktrialonline.com
Demo video: YouTube walkthrough

It's free to sign up and run several full trials without paying. I'm a solo developer building this actively — feedback from anyone (devs, lawyers, or just people curious about legal AI) is genuinely welcome.

How I Made AI Behave Differently Based on Conversation Context (Multi-Role Prompt Engineering)

Jesper Deng — Sat, 09 May 2026 08:35:09 +0000

I've been working on a project that requires multiple AI "characters" to behave differently in the same conversation — think of it like NPCs in a game, except each one needs to respond based on their role, personality, and the current situation.

Here's what I learned about making this work reliably.

The problem

If you just tell an AI "you are character A" and "you are character B" in separate prompts, they all end up sounding the same. Generic. Helpful. Boring. You need them to have distinct behaviors — one should be cooperative, another defensive, another authoritative.

What actually works

1. Behavioral constraints > personality descriptions

Bad:

You are a friendly witness who is helpful.

Better:

You are a witness being questioned. Rules:
- Only answer what is directly asked
- If the question is vague, ask for clarification
- Never volunteer extra information
- If pressed on a contradiction, become defensive

Constraints produce more consistent behavior than adjectives.

2. Context-dependent behavior switching

The same character might need to behave differently depending on who's talking to them. I handle this by passing a mode parameter in the system prompt:

function buildPrompt(character: Character, mode: 'friendly' | 'hostile') {
  const base = `You are ${character.name}. Background: ${character.bio}`;

  if (mode === 'friendly') {
    return `${base}\n\nBehavior: Be cooperative. Give detailed answers. Expand on your responses when appropriate.`;
  } else {
    return `${base}\n\nBehavior: Be defensive. Give minimal answers. Only confirm what you cannot deny. Redirect when possible.`;
  }
}

This tiny switch makes a huge difference in how natural the responses feel.

3. State management across turns

The hardest part: making characters remember what happened earlier and adjust. I maintain a simplified state object:

interface ConversationState {
  currentSpeaker: string;
  previousStatements: string[];
  contradictions: string[];
  mood: 'neutral' | 'defensive' | 'confident';
}

Before each AI call, I inject a summary of what's happened so far. This keeps responses contextually aware without blowing up the token count.

4. The "don't be helpful" problem

LLMs are trained to be helpful. When you need a character to be evasive or unhelpful, you have to fight against this training. What works:

Explicitly say "you do NOT want to help the questioner"
Give the character a motivation for being difficult
Add examples of deflection in the prompt

You are being questioned about something you want to hide.
Your goal is to answer without revealing [specific fact].
Techniques you use: giving technically true but misleading answers,
answering a different question than what was asked, saying "I don't recall."

5. Temperature matters more than you think

For authoritative characters (judges, experts), use lower temperature (0.3-0.5). They should be consistent and decisive.

For emotional or unpredictable characters, bump it up (0.7-0.9). The randomness makes them feel more human.

Key takeaway

Multi-role AI isn't about writing better character descriptions. It's about defining behavioral rules, injecting context, and fighting the model's default "helpful assistant" mode. Once I figured that out, everything clicked.

Would love to hear if anyone else is doing multi-agent stuff — what patterns are you using?

Forem: Jesper Deng