Monolithic AI
- One model
- Too much context
- Too many tools
- Too much responsibility
Architecture
The Progenitor System
The multi-agent architecture behind Progny.
Progny comes from Progenitor: the source, the origin, the parent system from which intelligent workflows begin. The Progenitor System creates, coordinates, and retires temporary AI agents so complex work can be handled by many focused specialists instead of one overloaded model.
AI is not the center of Progny. People are.
Why this architecture matters
Artificial intelligence is becoming more capable, but also more complex. Many systems respond by giving one model more context, more tools, more permissions, and more responsibility.
That creates systems that are expensive, difficult to govern, harder to observe, and more likely to use irrelevant information.
Intelligence should be decomposed rather than accumulated.
Instead of asking one increasingly complex AI to solve everything, the architecture divides work into many smaller tasks handled by temporary, specialized agents.
Progenitor
- One objective
- Many specialists
- Scoped context
- Clean termination
Why Progny uses the Progenitor
Progny is built around living profiles, verified information, and continuous human context. Preparing someone for an interview, verifying experience, organizing a profile, analyzing career history, generating documents, and researching companies are fundamentally different tasks.
The user experiences one assistant. Behind the scenes, many temporary specialists may have collaborated.
User asks
Prepare me for an interview at this company.
Profile Agent
Company Research Agent
Evidence Agent
Matching Agent
Interview Agent
Writer Agent
Why the name Progny?
The name Progny comes from Progenitor. A progenitor is an origin, an ancestor, or the source from which something begins.
The original Progenitor System was conceived during the development of ForteAI as an orchestration architecture for coordinating specialized AI workers across complex recruitment workflows.
At first, it solved a technical problem: how to create intelligence that could scale by coordinating many focused agents instead of relying on one overloaded system.
Over time, that idea became bigger. Technology is not the origin of progress. People are.
Progenitor
Progny
Core principles
Dynamic Agent Provisioning
Why should you care?
Permanent AI workers accumulate unnecessary responsibility and context over time.
The Progenitor creates agents only when work exists. Each receives a narrow objective, temporary context, limited permissions, required tools, memory boundaries, and execution constraints.
Hierarchical Orchestration
Why should you care?
Large problems become easier to solve when they are broken into smaller ones.
The Progenitor acts as the parent system. It decomposes work, creates agents, routes tasks, monitors execution, validates outputs, and combines results.
Ephemeral Intelligence
Why should you care?
Temporary workers create fewer long-term problems than permanent ones.
Agents are created for one purpose. They execute, return results, and disappear. Persistent knowledge belongs to the platform, not individual agents.
Parallel Execution
Why should you care?
Independent work should happen simultaneously whenever possible.
The Progenitor distributes work across specialized agents operating in parallel. This supports lower latency, workload isolation, and efficient resource use.
Context Isolation
Why should you care?
An AI should not know everything simply because it can.
Each agent receives only the information required for its task. It does not inherit unrelated conversations, documents, global memory, or permissions.
Capability Injection
Why should you care?
Most tasks do not require every tool.
The Progenitor dynamically provides only the models, tools, retrieval systems, APIs, memory scopes, and compute resources needed for the objective. Nothing more. Nothing less.
Agent lifecycle
Lifecycle management prevents uncontrolled execution, orphaned agents, and unnecessary resource consumption while making the system predictable and observable.
Objective received
Task planning
Agent creation
Context injection
Permission assignment
Execution
Validation
Aggregation
Termination
Temporary agent retires.
The orchestration engine
The orchestration engine functions like the operating system of the multi-agent environment. Individual agents perform work. The Progenitor coordinates the ecosystem.
Parent system
Progenitor
Objective decompositionDependency analysisSchedulingTask routingResource allocationRetry policiesTimeout handlingFailure recoveryConflict resolutionExecution monitoringResult aggregation
Memory architecture
Agents retrieve from persistent knowledge but never permanently own it.
Working Memory
Temporary context for one agent. Destroyed after execution.
Shared Task Memory
Information exchanged inside one workflow. Removed when the workflow completes.
Persistent Knowledge
Databases, documents, indexes, graphs, and historical records the platform owns.
Scoped, observable, governed
Security follows least-privilege execution. Each agent receives temporary credentials, scoped permissions, isolated execution boundaries, restricted tool access, and limited memory visibility.
Every stage can be observed: lifecycle events, execution duration, tool usage, API calls, resource consumption, retries, failures, and outputs.
Scoped accessTemporary credentialsTool limitsExecution logsRetry policyOutput validation
Failure stays local
In monolithic AI systems, one failure can affect the entire workflow. In the Progenitor System, failures are isolated to individual agents.
If one agent fails, the system can restart it, replace it, reassign the task, or ignore the failure if it is non-critical. Other agents continue operating.
Profile
Evidence
Company
Matching
Writer
Replacement
Human state as context
Most AI systems reason over prompts, documents, or chat history.
Progny is different.
The Progenitor coordinates agents around structured, continuously evolving human state. Each agent receives only the information required for its objective.
The human is the source, not the AI.
Source
Human
Verified profile informationCareer historyProjectsEducationEvidence-backed claimsUser permissionsTimeline contextTask-specific intent
Orchestration is intelligence
The Progenitor System views intelligence not as a single model, but as an orchestrated ecosystem.
By separating orchestration from execution, persistent knowledge from temporary reasoning, and long-term human state from short-lived computation, the Progenitor System gives Progny a foundation for AI that is modular, scalable, observable, resilient, secure, and designed to evolve alongside people.
Progny is not trying to replace people with AI. Progny is building AI infrastructure that works for people.