Research
How Pulse AI separates product documentation from deeper research into agent work, review gates, and correction loops.
Last reviewed
The promise
Research explains why Pulse AI works the way it does. Product docs tell you how to operate the product; research gives the theory behind agent coordination, correction persistence, review gates, and autonomous workforce design.
Pulse AI keeps research separate from task docs so setup and support pages stay practical while papers can go deeper.
Research versus product docs
| Surface | Best use |
|---|---|
| Docs | Canonical product behavior, feature references, settings, and workflows. |
| Help Center | Problem-first recovery when something is blocked or confusing. |
| Academy | Structured lessons and practice paths. |
| Research Library | Deeper arguments about agent behavior, orchestration, safety, and evidence systems. |
Start with Docs when you need to operate Pulse AI today. Use Research when you want to understand the design logic behind systems like review, rubrics, Knowledge Items, or dispatch.
How research connects to proof
Research can inform review standards, but it is not task evidence by itself. A paper may explain why review gates exist; the task still needs its own current proof.
For example, a research gap map can say a docs route needs expansion, but the completed rewrite still needs:
- updated source content,
- screenshots or media where useful,
- a KI or deliverable package,
- validation against the live route,
- and a review verdict tied to the task.
Research gives context. The completion gate and Review Queue decide whether a specific piece of work is ready.
Example: using research responsibly
Suppose a reviewer asks why Pulse AI keeps corrections in rubrics instead of leaving them in chat. A research paper can explain correction persistence and agent behavior. The product proof still comes from the live rubric file, task closeout, and review history.
That distinction keeps public docs truthful. We can teach the reasoning without exposing agent governance material or pretending a theory paper proves a shipped feature.
Research visuals
Research media should help readers understand the model:
| Media | Use |
|---|---|
dispatch-workflow.svg | Explain orchestration and dispatch flow. |
role-orchestration-model.svg | Show how roles, skills, and review connect. |
strict-lenient-skill-matrix.svg | Explain why some proof gates are strict. |
| Product screenshots | Use only when the paper refers to a current visible feature. |
The current public research assets live under /research/, including diagrams such as /research/dispatch-workflow.svg.
Research library launch state
The browsable Research Library is not open on the current public launch. This guide is the public operating reference for how research relates to product truth and task proof; do not send users to the gated library route.
The maintained research program covers:
- Machine Psychology: agent behavior, correction patterns, and role adoption fidelity.
- Orchestration Theory: multi-agent coordination, dispatch, and skill frameworks.
- Safety & Ethics: scope containment, correction persistence, and audit infrastructure.
Pulse AI will add a public library link here only after the browse, filter, paper, and recovery routes are enabled and verified together.
Recovery
If a paper and the current product disagree, treat the paper as historical context and verify the live source, release evidence, or runtime before updating product docs. Record the mismatch so the research library can be revised without converting theory into a shipping claim.
If this guide or a linked research asset fails to load, use the public Docs and Help surfaces while reporting the broken route through support with a redacted screenshot and the page URL.