Cognaptus Insights

General-purpose language models can solve math puzzles and explain Kant, but struggle to identify a ticker or classify earnings tone. What the financial world needs isn’t more reasoning—it’s better reading. Over the past year, large language models (LLMs) have surged into every corner of applied AI, and finance is no exception. But while the promise of “reasoning engines” captivates headlines, the pain point for financial tasks is much simpler—and more niche. ...

Traces of War: Surviving the LLM Arms Race The AI frontier is heating up—not just in innovation, but in protectionism. As open-source large language models (LLMs) flood the field, a parallel move is underway: foundation model providers are fortifying their most powerful models behind proprietary walls. A new tactic in this defensive strategy is antidistillation sampling—a method to make reasoning traces unlearnable for student models without compromising their usefulness to humans. It works by subtly modifying the model’s next-token sampling process so that each generated token is still probable under the original model but would lead to higher loss if used to fine-tune a student model. This is done by incorporating gradients from a proxy student model and penalizing tokens that improve the student’s learning. In practice, this significantly reduces the effectiveness of distillation. For example, in benchmarks like GSM8K and MATH, models distilled from antidistilled traces performed 40–60% worse than those trained on regular traces—without harming the original teacher’s performance. ...

The Crossroads of Reason: When AI Hallucinates with Purpose On this day of reflection and sacrifice, we ask not what AI can do, but what it should become. Good Friday is not just a historical commemoration—it’s a paradox made holy: a moment when failure is reinterpreted as fulfillment, when death is the prelude to transformation. In today’s Cognaptus Insights, we draw inspiration from this theme to reimagine the way we evaluate, guide, and build large language models (LLMs). ...

The next generation of quantitative investment agents must be more than data-driven—they must be logic-aware and structurally adaptive. Two recently published research efforts provide important insights into how reasoning patterns and evolving workflows can be integrated to create intelligent, verticalized financial agents. Kimina-Prover explores how reinforcement learning can embed formal reasoning capabilities within a language model for theorem proving. Learning to Be a Doctor shows how workflows can evolve dynamically based on diagnostic feedback, creating adaptable multi-agent frameworks. While each stems from distinct domains—formal logic and medical diagnostics—their approaches are deeply relevant to two classic quant strategies: the Black-Litterman portfolio optimizer and a sentiment/technical-driven Bitcoin perpetual futures trader. ...

Crunch Time for AI: Photonic Chips Enter the Menu In the diet of modern artificial intelligence, chips are the staple. For decades, CPUs, GPUs, and more recently TPUs, have powered the explosion of deep learning. But what if the future of AI isn’t just about faster silicon—it’s about harnessing the speed of light itself? Two recent Nature papers—Hua et al. (2025)1 and Ahmed et al. (2025)2—offer a potent answer: photonic computing is no longer experimental garnish—it’s becoming the main course. ...

When your AI believes too much, you pay the price. AI-driven quantitative trading is supposed to be smart—smarter than the market, even. But just like scientific AI systems that hallucinate new protein structures that don’t exist, trading models can conjure signals out of thin air. These errors aren’t just false positives—they’re corrosive hallucinations: misleading outputs that look plausible, alter real decisions, and resist detection until it’s too late. The Science of Hallucination Comes to Finance In a recent philosophical exploration of AI in science, Charles Rathkopf introduced the concept of corrosive hallucinations—a specific kind of model error that is both epistemically disruptive and resistant to anticipation1. These are not benign missteps. They’re illusions that change the course of reasoning, especially dangerous when embedded in high-stakes workflows. ...

“Every technical revolution rewires the old system—but does it fry the whole board or just swap out the chips?” The enterprise tech stack is bracing for another seismic shift. At the heart of it lies a crucial question: Can today’s emerging AI models—agentic, modular, stream-driven—peacefully integrate with yesterday’s deterministic data flows, or will they inevitably upend them? The Legacy Backbone: Rigid Yet Reliable Enterprise data architecture is built on linear pipelines: extract, transform, load (ETL); batch jobs; pre-defined triggers. These pipelines are optimized for reliability, auditability, and control. Every data flow is modeled like a supply chain: predictable, slow-moving, and deeply interconnected with compliance and governance layers. ...

In the wild, survival doesn’t require you to outrun the lion—it just means outrunning the slowest gazelle. Surprisingly, this logic also applies to business strategy. When we introduce AI into business decision-making, we’re not just dealing with isolated optimization problems—we’re engaging in a complex game, with rivals, competitors, and market players who also make moves. One key trap in this game is assuming that opponents are perfect. That assumption sounds safe—but it can be paralyzing. ...

In a world awash with data and decisions, the tools we use to think are just as important as the thoughts themselves. That’s why the Dual Engines of Thoughts (DEoT) framework, recently introduced by NeuroWatt, is such a game-changer. It’s not just another spin on reasoning chains—it’s a whole new architecture of thought. 🧠 The Problem with Single-Track Thinking Most reasoning systems rely on either a single engine (a one-track logic flow like Chain-of-Thought) or a multi-agent setup (such as AutoGen) where agents collaborate on subtasks. However, both have trade-offs: ...

The Invisible Hand of the Algorithm You open your favorite map app and follow a suggestion for brunch. So do thousands of others. Without realizing it, you’ve just participated in a city-scale experiment in behavioral automation—guided by a machine learning model. Behind the scenes, recommender systems are not only shaping what you see but where you physically go. This isn’t just about convenience—it’s about the systemic effects of AI on our cities and social fabric. ...