Overnight Research Cycles

Overview

Overnight Research Cycles are autonomous research experiments that Praxiom runs while you sleep. Each cycle explores different angles across your research sources, scores the results by quality (RQS), and keeps only the high-quality insights. In the morning, you receive a digest summarising what was discovered.

---

How It Works

1. Scheduling — Cycles run automatically at 2 AM UTC every night when enabled
2. Experimentation — Each cycle runs 8–15 experiments, each exploring a different research angle derived from your Research Program
3. Quality scoring — Every experiment is scored by the Research Quality Score (RQS) system
4. Curation — Only high-quality insights (above the RQS threshold) are accepted into your workspace
5. Digest — A morning summary email lists what was discovered, with the best RQS score and accepted count

Credit Cost

Each overnight cycle consumes approximately 8 agent credits. The exact cost depends on the number of experiments and their complexity.

---

Requirements

All three conditions must be met before cycles can run:

Requirement	Details
Power plan	Only available on the Power tier
3+ research sources	Your workspace needs at least 3 uploaded sources to generate meaningful experiment angles
Research Program defined	You must configure at least one goal in the Research Program editor

If any requirement is not met, the toggle in Settings shows a notice explaining what is missing.

---

Enabling Overnight Cycles

Navigate to Settings → Overnight Cycles to find the controls:

Master Toggle

Enable overnight research cycles — Turns the nightly experiment scheduler on or off. When enabled, cycles begin the next night at 2 AM UTC.

Morning Digest Email

Send morning digest email — When enabled, workspace admins receive an email each morning after a cycle completes, summarising:

• Number of experiments run
• Number of insights accepted (passed RQS threshold)
• Best RQS score achieved
• A brief digest summary of findings

---

Research Program

The Research Program defines what overnight cycles should explore. Configure it in Settings → Research Program.

Goals

Define what you want to learn. Goals guide the angle generation for each experiment. Examples:

• "Understand why users abandon the onboarding flow"
• "Identify competitive differentiators in our market"

Hypotheses

State what you believe to be true. The agent will design experiments that test or challenge these hypotheses:

• "Users drop off because the setup wizard has too many steps"
• "Enterprise users need SSO before they will adopt"

Anti-Goals

Explicitly exclude topics you do not want explored:

• "Do not research pricing strategies"
• "Ignore mobile-specific issues"

Focus Areas

Narrow the scope to specific domains or themes:

• "Checkout and payment experience"
• "Developer onboarding and API documentation"

---

Viewing Cycle History

After cycles run, view the results in Settings → Overnight Cycles → Cycle History.

Cycle List

GET /api/workspaces/{workspace_id}/cycles

Each cycle shows:

Field	Description
status	running, completed, or failed
experiment_count	Total experiments in this cycle
accepted_count	Experiments that passed the RQS threshold
best_rqs	Highest RQS score achieved
digest_summary	Brief text summary of findings
started_at	When the cycle began
completed_at	When the cycle finished

Cycle Detail

GET /api/workspaces/{workspace_id}/cycles/{cycle_id}

Drill into a specific cycle to see each experiment:

Field	Description
angle	The research angle explored
rqs_score	Quality score for this experiment
accepted	Whether the results were accepted
insight_count	Number of insights generated

---

What's Next

• Dashboard Quality — view RQS trends and quality breakdown
• Billing & Plans — Power plan details and credit costs
• Chat Agent — deep research via the /research command