How It Works

Sparse Halo changes shape with the problem.

The main workspace shifts between direct Chatbot conversation and structured Cabinet rigor, while SAT Prep stays separate for adaptive Digital SAT practice. Each surface is designed around the shape of the work instead of forcing everything through one interface.

Core Logic

The product family is built around one simple idea: the right surface depends on the shape of the work.

Sparse Halo keeps the main workspace focused on two modes: Chatbot for direct continuity and Cabinet for structured multi-agent rigor. SAT Prep (Beta) sits beside that workspace as a separate adaptive lane for Digital SAT practice. The point is not to add complexity for its own sake. The point is to give different kinds of work a shape that actually fits.

Main Workspace

Chatbot + Cabinet

Separate Lane

SAT Prep (Beta)

Session Model

Temporary by default, local only when you choose

The short version

Start Direct

Use Chatbot when you want a clean, ongoing conversation and the fastest path to a useful reply.

Add Pressure

Move to Cabinet when the problem needs critique, planning, multiple perspectives, or a stronger synthesis.

Switch Tracks

Use SAT Prep when the work is adaptive Digital SAT practice, scoring, and guided academic feedback rather than open-ended workspace reasoning.

Keep Retention Deliberate

Keep sessions temporary unless you explicitly decide to save them locally on your own device.

Start in Chatbot

Chatbot is the direct workspace mode. You ask a question, stay in one live thread, and can switch models for future replies without rewriting everything that already happened.

It is built for continuity: follow-up questions, reframing, drafting, iteration, and the kind of back-and-forth where the thread itself matters as much as the latest answer.

Escalate to Cabinet

Cabinet is the structured workspace mode. You configure a 2-5 agent panel, shape how the discussion opens, and let a synthesis model consolidate the strongest final answer after the round-by-round exchange.

This is the mode for prompts that are not really single-answer problems: plans with tradeoffs, ideas that need pressure-testing, or situations where the first instinct is probably incomplete. Judge scoring and optional web grounding can add even more rigor when the task calls for it.

Use SAT Prep Separately

SAT Prep (Beta) is not a third workspace mode. It is a separate adaptive Digital SAT surface with its own 20-question diagnostic, targeted practice flow, three advisor lanes, and score estimate.

That separation is deliberate. Academic practice has different pacing, scoring, and feedback needs than Chatbot or Cabinet, so Sparse Halo gives it a dedicated lane instead of forcing it through the main workspace.

Keep Sessions Temporary

Sparse Halo's privacy model is about default retention. Chatbot and Cabinet each keep their own active thread in the browser, and starting fresh clears the active workspace instead of turning every interaction into a permanent archive by default.

If you want to keep something, local saving is opt-in on your own device. That retention model should not be confused with a claim that outside model providers never process requests in order to generate replies. The privacy policy explains the details.

Move Between Modes On Purpose

Sparse Halo supports explicit carry-context handoff between Chatbot and Cabinet. You can begin with a direct conversation, switch into a more rigorous multi-agent pass, and bring the right context over deliberately.

That separation is important. Chatbot and Cabinet stay legible inside the main workspace, while SAT Prep remains separate by design. Each surface keeps its own controls, pacing, and role, which makes the product easier to understand than a single interface trying to do everything at once.