🕒 Now - What I'm Currently Focused On

Mar 26, 2025·
Chris Mascioli
Chris Mascioli
· 2 min read

Welcome to my “Now” page 👋

Table of Contents

Current Research Focus

I’m currently working on several projects at the intersection of AI, transformers, LLMs, and multi-agent systems:

1. Generalizable TRON Agents with Transformer Backbones

I’m extending PyMarketSim and the TRON (Trained Response Order Network) agent architecture to incorporate transformer backbones. The goal is to create agents that can generalize across different market environments, rather than requiring a separate TRON agent to be trained for each specific market configuration.

This research aims to significantly reduce the training overhead while enabling more robust trading strategies that can adapt to changing market conditions. By leveraging the contextual understanding capabilities of transformers, these agents should be able to recognize patterns across different market scenarios.

2. LLMs as Bargaining Agents

I’m investigating the use of Large Language Models (LLMs) as bargaining agents in strategic interactions. This work includes:

  • Developing frameworks to evaluate LLM performance in bargaining scenarios
  • Testing different prompting strategies to improve negotiation capabilities
  • Comparing LLM performance to traditional game-theoretic equilibrium predictions
  • Exploring how LLMs might be used in automated negotiation systems

This research has applications in market design, automated negotiation systems, and understanding the strategic capabilities of foundation models.

3. Mechanism Design Approaches to Mixture of Experts

I’m developing new approaches to Mixture of Experts (MoE) architectures using ideas from mechanism design and economics. This interdisciplinary work explores:

  • Incentive-compatible routing mechanisms for expert selection
  • Market-based approaches to expert specialization and load balancing
  • Applying concepts from social choice theory to improve expert consensus methods
  • Creating more efficient and effective models by drawing on economic principles

This research aims to address current inefficiencies in MoE architectures by bringing in theoretical insights from decades of research in economic mechanism design.

Why a “Now” Page?

A “Now” page is a snapshot of my current professional focus and interests. Unlike social media updates or blog posts that can get buried in feeds, this page stays in one place and is regularly updated.

Key benefits:

  1. It helps me clarify my priorities and commitments
  2. It provides an easy way for colleagues and friends to see what I’m up to
  3. It creates a record of my professional journey over time

Recent Updates

  • March 2025: Updated with current research projects
  • March 2025: Initial creation of Now page

This page was inspired by Derek Sivers’ now page movement.

Last updated: March 26, 2025

Now Projects