Onyx LEGION ← Back to App

Getting the Most from Legion

Your tokens are valuable. This guide ensures every query delivers real value.

📖 5 minute read

💰 Your Token Economy

🏛️

Legion Mode

~600 tokens

5 AI models debate across 4 rounds, then synthesize the best answer. This is what you came for.

Parallel Mode

~200 tokens

All 5 models answer simultaneously. You see responses side-by-side. Good for quick comparisons.

💬

Single Mode

~40 tokens

One model, one answer. Not the best use of Legion unless you specifically need a particular model.

💡 Bottom line: If you're using Single mode, you're not using Legion's superpower. Save that for when you specifically need Claude's writing style or GPT's code generation.

🧠 How Legion Mode Works

💭

Round 1

Initial Response

Each AI answers your question independently

⚔️

Round 2

Cross-Examination

Models see and critique each other's answers

🔄

Round 3

Refinement

Models improve based on feedback

Round 4

Synthesis

Best ideas combined into one answer

The result isn't just "what Claude thinks" or "what GPT thinks" — it's the combined intelligence of five different AI architectures, stress-tested against each other.

✍️ Writing Great Questions

Do This

  • Give context

    Your situation, constraints, what you've already tried

  • Be specific with numbers

    Budget, timeline, team size, metrics

  • Ask for tradeoffs

    "What are the pros and cons" beats "what should I do"

  • Frame decisions

    "Choosing between A and B given X constraints"

Don't Do This

  • Simple factual queries

    "What year was Tesla founded?" — Google this instead

  • One-word questions

    "Bitcoin?" — AI needs context to help you

  • Objectively correct answers

    "What is 2+2?" — No debate needed here

  • Small talk

    "How are you?" — Save tokens for real questions

📝 Example Prompts (Steal These)

IT Architecture

We're migrating from a monolith to microservices. Should we go with Kubernetes on GCP, ECS on AWS, or stay with managed containers on Azure? Team of 8 engineers, $400k annual cloud budget, need to maintain SOC 2 compliance. What are the tradeoffs?

Strategy

Our competitor just launched an AI feature. Do we build our own, buy/integrate a vendor solution, or partner with an AI startup? We have 6 months runway before market pressure hits. Engineering team of 15, $500k budget for this initiative.

Technical Research

Evaluate the viability of RAG vs fine-tuning for a domain-specific enterprise AI assistant. Consider cost, latency, accuracy, and maintenance burden. Our domain is financial compliance with ~50k documents.

Career

I have offers from a FAANG company ($220k) and a Series B startup ($160k + 0.8% equity, last valued at $50M). 10 years experience, want to be VP Engineering within 5 years. How should I evaluate this decision?

M&A / Investment

Assess the risks of acquiring a 20-person AI startup with strong tech but no revenue vs building an internal AI team over 18 months. $5M budget either way. We're a $100M ARR B2B SaaS company.

🎯 When to Use Each Mode

Use Case Legion Parallel Single
High-stakes decisions ✓ Best
Complex tradeoff analysis ✓ Best
Research questions ✓ Best Good
Comparing approaches quickly ✓ Best
Creative brainstorming ✓ Best
Need Claude's writing specifically ✓ Best
Need GPT for code ✓ Best
Testing a prompt before Legion ✓ Best

💡 Pro Tips

🔍 Search your history

Full-text search across all your conversations. Build a personal knowledge base of AI-vetted decisions.

📥 Export everything

Copy, download as Markdown, or share any response. Great for documentation.

⌨️ Keyboard shortcuts

Enter to send, Shift+Enter for new line. Speed up your workflow.

🔑 BYOK for power users

Use your own API keys for unlimited queries without spending tokens.

🔄 Wallet not updating?

After purchasing tokens, refresh your browser or log out and back in to see your updated balance.

Ready to ask something important?

You now know how to get maximum value from Legion.

Open Legion