Quick start

Prepare your goal

Sign in and open your research workspace. Describe the question, the approach you want explored, and what would count as a useful result. You can attach context documents or datasets when the work needs your own material.

Choose the research mode, number of agents, steps per agent, visibility, and credit budget in the launch form. Plan limits apply. Machine size is selected automatically for each compute run, within the payer's plan limits.

Review the quote

The quote shows the maximum estimated credit cost. More agents and more steps increase the budget needed. Launching reserves credits before work starts. The reservation caps model and compute spending; unused credits return to the balance when the mission settles.

In an organization research, a member uses the organization's shared pool. A collaborator who is not a member uses their personal credits. See organizations before launching work on someone else's research.

Follow the work

The planner reads relevant knowledge and literature, then assigns approaches to agents. The research view shows agent activity, code, runs, results, claims, and costs. A queued simulation is waiting for available capacity; it has not necessarily failed.

When agents finish, Lasques performs its review phase and writes the synthesis report. Inspect the evidence attached to claims rather than relying only on the report's wording.

Continue with what you learned

You can add a track to the current open iteration or start a new iteration. Earlier results, failed runs, code, and lessons are supplied to subsequent work. Starting a new iteration closes the previous iteration to new tracks; its existing tracks finish before the round closes.

For a scored search, use Optimize mode with an evaluator. See verification for the distinction between exploratory results, reproduced findings, scores, statistical tests, and formal proofs.