Guide

Drop in your problem and an agent team analyzes, researches, and writes — producing a document you can actually submit.

Quick Start

Enter your problem in the workspace and you're off. Answer a few questions and the agent team works in parallel to finish a draft.

1

Enter your problem → analysis skeleton in 30 seconds

2

Answer 2-3 questions → agent team auto-assembles

3

Team works in parallel → approve/reject results

4

Draft assembly + decision-maker simulation

Go to workspace

Agent Team

17 specialized agents, each with their own methodology. They level up and learn your preferences as you use them.

Research: Hayoon → Daeun → Doyoon
Strategy: Jungmin → Hyunwoo → Seunghyun
Execution: Seoyeon · Gyumin · Hyeyeon · Sujin · Minseo · Junseo · Yerin
Validation: Donghyuk · Jieun · Yunseok

Tip: Approving/rejecting agent output earns XP and levels them up. From Lv.2 they learn your patterns and outputs start to shift.

Unlocks: Some agents (like Doyoon, Seunghyun) unlock automatically once earlier agents in the chain hit a task threshold.

Agent hub

Detailed Tools

Outside of Quick Start, each stage can be used standalone. When run in order, context carries through.

Context chain: Score Reading (problem analysis) → Arrangement (execution design) → Rehearsal (pre-validation) → Ensemble (feedback) → Synthesis (integration). Context flows through.

Boss Simulator

Set up a real-life boss's personality using MBTI + birth chart, and practice reporting to them. The boss's mood shifts in real time, and input hints coach you along.

1

Enter the boss's MBTI + date of birth

2

Describe the situation (report, proposal, conflict, etc.)

3

Start the conversation — the boss's mood reacts live

4

Ends when you succeed at persuading or reach a conclusion

Boss Simulator
01

Score Reading | Reframe

When you get a task, you want to jump straight into execution — but if the premise is wrong, the whole report is wasted. Here you inspect the hidden assumptions behind the task and redefine the real question you should be answering.

How to use

1

Enter the task's context (origin, success criteria, stakeholders)

2

Surface the hidden premises together

3

Evaluate each premise yourself — likely true / uncertain / doubtful

4

The question gets redefined based on your evaluation

5

Pick one of several directions or write your own

Tip: Evaluating premises is the crux. The more you mark as "doubtful," the more fundamentally the reframing shifts direction.

Start
02

Arrangement | Recast

If you delegate everything to AI, the direction drifts. Here you decide what AI handles and what humans own — and you actually make the calls that need human judgment.

How to use

1

Pick the output type, team size, timeline, and enter the goal

2

The workflow is designed (governing idea, storyline, step-by-step roles)

3

Click each step to guide AI, or enter decisions where human judgment is needed

4

Adjust role assignment (AI / human / collab) and checkpoints

5

Verify the premises and critical path, then confirm

Tip: On human/collab steps, click the option chips to decide quickly. On AI steps, enter the direction or constraints you want AI to focus on.

Start
03

Rehearsal | Pre-validation

The same report needs different framing for the CEO vs. the working team. Simulating key stakeholder reactions before sending lets you shore up weak spots in advance.

How to use

1

Set up stakeholder personas (from presets or custom)

2

Enter the material for review (can carry over from the previous step)

3

Simulate feedback from that persona's perspective

4

Review failure scenarios and the 3 risk categories — core threat, manageable, unspoken risk

5

Check approval conditions and decide how to address weak spots

Tip: Watch for "unspoken risks." The problems everyone sees but no one raises are often the most dangerous.

Start
04

Ensemble Practice | Refine

A single pass won't produce a perfect result. Iterate with stakeholder feedback — each loop accumulates context and sharpens the output.

How to use

1

Enter the feedback from rehearsal

2

Feedback is applied and the analysis improves

3

Check the convergence rate — once converged enough, you're ready to execute

4

Iterate further if needed (context compounds)

Tip: The point isn't "iteration" itself — it's context accumulation. What you discover in each iteration becomes a constraint for the next.

Start
05

Synthesis | Integrate perspectives

When you have multiple analyses or opinions, simply combining them buries what matters. Here you structure agreements and conflicts, then decide on each conflict yourself to land the final conclusion.

How to use

1

Enter the sources to compare (AI analyses, expert opinions, etc.)

2

Each source's core claims are extracted and organized

3

Agreements and conflicts are auto-categorized

4

Decide yourself which side to take on each conflict

5

The final synthesis document is generated

Tip: Coming from the previous step (Refine), context carries over automatically. When used standalone, paste the sources yourself.

Start