For more than a decade, Palantir has been one of the few companies to demonstrate that large‑scale knowledge graphs and ontologies can deliver real value in complex, mission‑critical environments. In intelligence, defense, energy and large industrial organizations, Palantir has shown what is possible when data, models and domain knowledge are brought together coherently.
At the same time, Palantir’s success has highlighted a hard truth: traditional ontology‑driven platforms are extraordinarily expensive, complex, and slow to deploy. As a result, they remain inaccessible to the vast majority of enterprises.
Geminos KnowledgeWay was designed to deliver the core benefits of Palantir‑style knowledge platforms—without the cost, rigidity, and time‑to‑value penalties that have historically limited adoption.
Palantir’s strength—and its bottleneck
At the heart of Palantir’s approach is a highly structured ontology. This ontology defines entities, relationships, constraints, and semantics for a given domain, enabling consistent reasoning, analytics, and decision‑making at scale.
When it works, the results are powerful. But building and maintaining such an ontology is a massive undertaking. It typically requires:
- Large teams of highly specialized ontology engineers
- Deep, ongoing involvement from subject matter experts
- Months or years of upfront modeling before value is realized
- Continuous manual maintenance as the business and data evolve
This makes Palantir deployments slow, expensive and highly bespoke. Costs routinely run into the tens of millions of dollars, and implementations are often justified only for the most strategic or regulated use cases.
The hidden rigidity of traditional ontologies
Another limitation of traditional ontology‑first approaches is rigidity.
Ontologies assume that knowledge can be fully specified upfront. In reality, enterprise knowledge is messy, incomplete, and constantly changing. New terminology emerges, exceptions accumulate, and informal practices coexist with formal processes.
In Palantir‑style systems, accommodating this kind of evolution often requires revisiting core ontology design decisions—a process that is slow, costly, and dependent on scarce expertise. As a result, many organizations limit the scope of their knowledge graphs, reducing their usefulness to everyday users.
Why most organizations never get there
Because of this cost and complexity, most organizations simply cannot justify a Palantir‑class deployment. Instead, they fall back on lighter‑weight alternatives such as dashboards, document search, or RAG‑based chatbots.
While cheaper, these tools lack the structural understanding required to answer complex questions, preserve context, or support consistent decision‑making across the enterprise.
KnowledgeWay: the same destination, a different path
KnowledgeWay takes a fundamentally different approach to building enterprise knowledge graphs.
Rather than starting with a comprehensive, manually designed ontology, KnowledgeWay uses LLM‑driven agents to extract entities, relationships, and supporting evidence directly from documents and other sources. Ontology emerges incrementally, shaped by real usage and curated over time.
This agent‑assisted, bottom‑up approach delivers several advantages:
- Orders‑of‑magnitude faster time to value
- Dramatically lower upfront cost
- The ability to tolerate ambiguity and evolution
- Knowledge graphs that reflect how organizations actually work
Crucially, KnowledgeWay still produces a true Enterprise Knowledge Graph—not just a graph‑shaped index to text. Entities and relationships are explicit, human‑readable and continuously improvable.
Human‑curated, not human‑built
In KnowledgeWay, humans remain in the loop—but in a fundamentally different role.
Instead of manually constructing the graph from scratch, knowledge engineers curate and refine a graph that agents have already created. Task agents surface potential issues such as synonym conflicts, disconnected clusters, or missing relationships, and suggest fixes.
Subject matter experts contribute where their expertise matters most: resolving ambiguity, validating assumptions, and capturing tacit knowledge that never appears in documents.
This shift—from hand‑built to human‑curated—dramatically reduces cost while improving coverage and realism.
Software 3.0 versus Software 1.0
Seen through the lens of Andrej Karpathy’s “Software 3.0,” the difference is stark.
Palantir is largely a Software 1.0 system: powerful, but heavily dependent on human‑written models and logic. KnowledgeWay is Software 3.0: agents and LLMs do most of the work, guided by natural language intent and refined by human judgment.
The result is a platform that scales with the organization, rather than one that requires ever‑larger teams to maintain it.
Enterprise‑grade, without enterprise drag
KnowledgeWay delivers many of the outcomes organizations seek from Palantir:
- Consistent, explainable answers grounded in enterprise knowledge
- Cross‑document reasoning and contextual understanding
- A shared semantic layer across teams and systems
But it does so with:
- Far lower total cost of ownership
- Faster deployment and iteration
- Support for any public or private LLM, including on‑prem models
- A clear path to continuous improvement
In short, KnowledgeWay is best understood as a next‑generation Palantir: a platform that brings enterprise knowledge graphs into the age of agents and LLMs, making them practical, scalable, and economically viable for a much broader set of organizations.
For enterprises that want the power of ontology‑driven intelligence without the traditional barriers, KnowledgeWay offers a new path forward.