Jumptuit Highlights at SCSP AI+ Expo May 7, 2026

The Special Competitive Studies Project (SCSP): A Non-Partisan U.S. Think Tank Founded by Former Google CEO Eric Schmidt

SCSP AI+ Expo 2026 Opening Remarks: Unveiling the Large Dynamic Reasoning Model: AI, the Physical World, and Human Decision Making. "Jumptuit presented their new AI model that forecasts how global events impact supply chains and industries. Using live data, this model predicts cascading risks to provide vital intelligence before major decisions are made."

Jumptuit Founder & CEO, Donald Leka:

"It’s great to be back at the AI+ Expo in Washington, D.C. A lot has happened in the world since our presentation last year.

In a little over two months, Operation Epic Fury fundamentally altered geopolitical dynamics from Southeast Asia and East Asia, to Africa and Europe, in the Middle East and even within the GCC itself – from alliances, to trade relations, to investment priorities. These geopolitical tectonic shifts and cross-regional and cross-sector impacts were not anticipated by existing forecasting systems and methods.

Today we will discuss not only how AI can help us anticipate events but also forecast how the dust will settle in the aftermath of a geopolitical event.

LLMs have proven inadequate to accurately assess and forecast events in the physical world. We will discuss how LLMs can be supplemented with the help of a new breakthrough AI technology.

We will introduce you to a new vision of Artificial Intelligence that can bridge the digital world with the physical world in new and unprecedented ways. So let’s get started…"

The State of Large Language Models (LLMs): Hallucination Rates

Jumptuit Founder & CEO Donald Leka at SCSP AI+ Expo 2026:

"What is the state of LLMs today in relation to forecasting, and understanding current events?

LLMs are not a good foundation for forecasting events...not just because you're dealing with static data, but because of hallucinations. Hallucinations are not minor errors, these are not random errors, these are systemic errors, and they lead in many cases to fabrications that we all have experienced using LLMs based on no data at all.

Whereas, LLMs are good for many things, forecasting events is not what they are designed to do."

How to Fix AI’s Data Problem

Jumptuit Founder & CEO Donald Leka at SCSP AI+ Expo 2026:

Data Pipeline Flow and Cross-Sector Analysis:

“Jumptuit starts with Live Homogeneous Vector Data Pipelines. These pipelines are organized by Aerospace and Defense, Food and Agriculture, Energy, Environmental for Physical Phenomena, Finance, Manufacturing, Metals and Mining, Public Health, Transport and Logistics, and many other vertical sectors.

All of that data flows into a Unified High-Dimensional Vector Space. This is where the cross-sector analysis occurs, and cross-sector risk indices are generated.

This is also where intelligent associations are created within the data itself – mathematical indicators, numerical vectors with magnitude and direction, benchmarks and thresholds, data correlations, and coupled variables. Meaningful, intelligent insights are formed here, stored in the system, and can be accessed when faced with future similar probable events.”

Data Constitution:

“Jumptuit focuses on observable phenomena and human activity, primarily quantitative data. Our data mix is 80% quantitative and 20% qualitative. We have connected to hyper-localized quantitative data sources all over the world.

In contrast, LLMs rely primarily on web crawlers to capture their content.

So the accuracy level is higher because we are looking at empirical data. The objectivity is greater because we are looking at things that are measurable and verifiable, as opposed to human and AI-generated content, which is to a large extent unverifiable. And the consistency in terms of the methodologies we apply – because it is quantitative data and from verifiable sources which are primarily observation-based – makes a huge difference.”

Triangulation:

“Jumptuit uses triangulation from multiple sensors and audited, verifiable data sources. When we constitute a picture of what is happening in relation to an event in the world, what we are looking at is multiple data sources to construct that picture.”

This ensures the picture is built from various independent perspectives rather than a single source. If a satellite sensor, a ground-level IoT sensor, and a financial ledger all point to the same event, the confidence interval increases significantly.

How to Fix AI’s Reasoning Problem

The Problem: The physical world is driven by deeply interconnected, complex systems that trigger cross-sector ripple effects across the global economy, infrastructure, and supply chains. However, traditional Large Language Models (LLMs) fail to map and forecast these intricate cross-sector collateral consequences.

The Solution: Jumptuit has developed a flexible reasoning process that dynamically shifts among sign, comparative, causal, and analogical reasoning. This allows the AI model to observe, analyze, and understand live, cross-sector data streams from the physical world with sophisticated cross-sector ontologies enabling it to detect novel, complex conditions and anticipate events.

By dynamically adapting to live inputs, these autonomous reasoning permutations mitigate systemic risk and reduce the costs of geopolitical and environmental instability, augmenting human decision-making and forecasting event shockwaves before they materialize.

Jumptuit Founder & CEO Donald Leka at SCSP AI+ Expo 2026:

“Jumptuit observes phenomena in the physical world. We track events, transitional events, and events in motion, as well as probable events. Because clusters of signals appear before an event occurs, Jumptuit captures these precursors, antecedent elements, and human activity—including human behavior and human systems.

We treat human behavior and human systems as external variables to be observed.

Jumptuit utilizes various reasoning types. Depending on the live data input, we apply different permutations of sign, comparative, causal, analogical, verbal, and temporal reasoning.

By doing so, Jumptuit observes, analyzes, and understands live cross-sector data, ensuring our reasoning permutations are dynamically responsive to live inputs.

Verbal reasoning is ultimately the final stage. Once a core data response is formulated based on the live data input, verbal text is generated to describe the findings.”

Jumptuit’s Large Dynamic Reasoning Model (LDRM):

Dynamic Observation: Phenomena, Events, Transitional Events, Probable Events, Human Systems, Human Behavior

Dynamic Reasoning Methods: Sign, Comparative, Causal, Analogical, Verbal, Temporal

Autonomous Reasoning Permutations Dynamically Respond to Live Inputs

Detect Novel and Complex Transitional Events Generate Continuous Dynamic Scenario Forecasting of Probable Events