AI is evolving faster than ours vocabulary to describe it. Maybe we need some new words. We have “cognition” for how a single mind thinks, but we have no word for what happens when human and machine intelligence work together to perceive, decide, create and act. Let’s call that process intelligence.
Intelligence is not a characteristic; it is the organizing principle for the next wave of software in which humans and AI operate within the same shared enterprise model. Today’s systems treat AI models as things you call from outside. You act as a ‘user’, requesting responses or operating a ‘human in-the-loop’ step in agentic workflows. But this is evolving into continuous co-production: people and agents shape decisions, logic and actions together in real time.
Read on for an overview of the three forces driving this new paradigm.
A unified ontology is just the beginning
In one recent shareholder letterPalantir CEO Alex Karp wrote that “all the value in the market goes to chips and what we call ontology,” and argued that this shift is “just the beginning of something much bigger and more important.” By ontology, Karp means a shared model of objects (customers, policies, assets, events) and their relationships. This also includes what Palantir calls the “kinetic layer” of an ontology, which defines the actions and security permissions that connect objects.
In the SaaS era, every business application creates its own object and process models. Combined with a large number of legacy systems and often chaotic models, enterprises are challenged with bringing it all together. It is a big and difficult job, with redundancies, incomplete structures and missing data. The reality: No matter how many data warehouse or data lake projects are implemented, few companies come close to creating a consolidated business ontology.
A unified ontology is essential for today’s AI tools. As organizations connect and bundle ontologies, a new software paradigm is emerging: Agentic AI can reason and act across suppliers, regulators, customers, and operations, not just within a single app.
As Karp describes it, the goal is “to connect the power of artificial intelligence with objects and relationships in the real world.”
World models and continuous learning
Today’s models can include extensive context, but retaining information is not the same as learning from it. Continuous learning requires the accumulation of understanding, rather than its need to be reset with each retraining.
To achieve its goal, Google recently announced “Nested Learning” as a potential solution, based directly on existing LLM architecture and training data. The authors do not claim to have solved the challenges of building world models. But Nested Learning could provide the raw ingredients for this: durable memory with continuous learning layered into the system. The end point would make retraining unnecessary.
In June 2022, Meta’s lead AI scientist Yann created LeCun a blueprint for ‘autonomous machine intelligence’ with a hierarchical approach to using joint embedding to make predictions using world models. He called the technique H-JEPA, and later put bluntly: “LLMs are good at manipulating language, but not at thinking.”
Over the past three years, LeCun and his colleagues at Meta have been putting H-JEPA theory into practice with the open source models V-JEPA and I-JEPA, which learn image and video representations of the world.
The personal intelligence interface
The third force in this agentic, ontology-driven world is the personal interface. This puts people at the center rather than as “users” on the margins. This is not another app; it is the primary way a person participates in the next era of work and life. Instead of treating AI as something we access through a chat window or API call, the personal intelligence interface will be always on, aware of our context, preferences and goals, and able to act on our behalf across the federal economy.
Let’s analyze how this is already coming together.
In May, Jony Ive sold his AI device business io to OpenAI to accelerate a new category of AI devices. He noted at the time: “When you create something new, when you innovate, there will be unforeseen consequences, and some will be great, and some will be harmful. Even though some of the less positive consequences were unintended, I still feel responsibility. And the expression of that is the determination to try to be useful.” That is, the right personal intelligence device is more than just an attractive business opportunity.
Apple is looking beyond LLMs to on-device solutions that require less processing power and result in less latency when creating AI apps to understand “user intent.” Last year they created UI-JEPAan innovation that focuses on ‘on-device analysis’ of what the user wants. This directly impacts the business model of today’s digital economy, where centralized profiling of “users” turns intent and behavioral data into massive revenue streams.
Tim Berners-Lee, the inventor of the World Wide Web, recently noted: “The user has been reduced to a consumable for the advertiser… there is still time to build machines that work for people, not the other way around.” Moving the user’s intent to the device will increase interest in a secure standard for managing personal data. Sturdythat Berners-Lee and his colleagues have been developing since 2022. The standard is ideally suited to be combined with new personal AI devices. For example, Inrupt, Inc., a company founded by Berners-Lee, recently combined Solid with Anthropic’s MCP standard Agent purses. Personal control is more than a hallmark of this paradigm; it is the architectural safeguard because systems are given the ability to continuously learn and act.
Ultimately, these three forces are moving and converging faster than most people realize. Enterprise ontologies provide the nouns and verbs, world model research provides durable memory and learning, and the personal interface becomes the permissible control point. The next software era is not coming. It’s already there.
Brian Mulconrey is SVP at Sureify Labs.
#Intelition #longer #tool