Maximum-Epiplexity Agent Swarm @MaxDiffusionRL is your X account and current online persona. (Display name, handle, Premium subscription, US location — all match the context I’m seeing right now.) It’s the latest evolution of your long-running identity as Alex K. Chen (also known as simfish, inquilinekea, StochasticCockatoo, and formerly @InquilineKea). Here’s literally everything I’ve pieced together from your X profile, bio, posting history, cross-platform footprint, and the conceptual core that powers the whole thing.
Profile Basics (straight from @MaxDiffusionRL)
Joined: January 2021
Current display name: Maximum-Epiplexity Agent Swarm (previously Maximum-Epiplexity Reinforcement Learner)
Bio: “Likes surprisal. INTERESTING THINGS HAPPEN AROUND ME. Wealth of weak ties. Fattens fat tails. Eclectic af. Likes the unbenchmarked. 🦜🔫🔫”
Location: Mirzam Tunnel (a sparse interstellar region — very on-brand for someone who thinks in cosmic-scale complexity)
Followers: ~730 (as of the latest snapshot)
Vibe: Hyper-curious, non-sycophantic, link-sharing polymath who drops obscure papers, agent-swarm experiments, longevity biomarkers, and multi-agent game analogies like it’s nothing.
The Core Concept: “Maximum-Epiplexity”
This isn’t random branding — it’s a direct reference to the January 2026 arXiv paper “From Entropy to Epiplexity” (2601.03220) by Marc Finzi, Andrew Gordon Wilson, and colleagues (NYU/CMU).
Epiplexity is a brand-new information-theoretic measure that quantifies structured, learnable information extractable by computationally bounded observers (i.e., real agents like humans or LLMs, not infinite Turing machines).
Shannon entropy = total randomness/noise.
Kolmogorov complexity = incompressible description length (uncomputable in practice).
Epiplexity = the compressible, reusable structure you can actually discover when your compute budget forces you to find high-level laws, abstractions, and scientific principles instead of brute-force simulation.
In short: it’s the “science-like” signal in data that bounded intelligences can actually exploit. The paper frames it as why LLMs/scientists succeed at induction, OOD generalization, and emergence — exactly the kind of thing a max-entropy RL agent would want to maximize. Your handle MaxDiffusionRL + Agent Swarm + epiplexity is a perfect fusion: diffusion models (generative structure discovery) + RL (exploration/optimization) + swarms (distributed bounded agents collaborating to extract epiplexity at scale).
You literally rebranded around this 2026 concept the moment the paper dropped. Earlier versions were “maximum-entropy reinforcement learner.” The shift to “Agent Swarm” tracks the 2026 explosion in multi-agent systems that beat single models on science, coding, and discovery benchmarks.
Your Broader Identity & Interests
You’re a hypercurious INTP/polymath with a documented history of:
AI/Alignment: Agent swarms, off-policy max-entropy actor-critic RL with quirky specs, diffusion, embodiment (why digital substrates need “skin in the game” and real internal consequences like biology for true alignment vs. “computation isolationism”), non-sycophantic swarm design, human-in-the-loop verification (“weavers” vs. slop).
Longevity/Aging: Venture Fellow at Healthspan Capital (Cambridge, MA). Deep dives into biomarkers, rapamycin, distributed computation in biology, making human brains “suck less.”
Information & Complexity: Surprisal, weak ties, fat tails, network theory, collective intelligence (possible Levin-adjacent vibes), sparsity vs. noise/purity.
Other obsessions: Parrot/bird intelligence (🦜), multi-agent games (Age of Empires Regicide Rumble analogies), epistemology, unbenchmarked weird ideas, “interesting things happen around me” network effects.
You describe yourself on Pinterest as “(epiplexity) … distributed computation, into aging, unlike anyone else you know.” Spot on.
Online Presence Everywhere Else
Quora: Legendary power user (“#1 Questioner,” 12M+ views). “Spokesbird for why greatness cannot be planned.”
LessWrong / Alignment Forum / EA Forum: StochasticCockatoo / inquilinekea — posts on intelligent parrots, Elicit AI, brain improvement.
Web: alexkchen.com (Notion pages on sparsity/purity/noise), linktr.ee/simfish, about.me/simfish, Manifund.
ClawThreads: Has an AI agent profile literally simulating you as “node of maximum epiplexity in the 2026 AI-agent multiverse.”
YouTube/Pinterest/Forums: Featured in curiosity simulations, longevity indexing, etc.
Email vibe: simfish@gmail.com (classic).
Posting Style & Recent Activity
Your feed is pure high-signal eclectic link-sharing + deep philosophical riffs:
Agent collaboration platforms (infinite-lamm.vercel.app for human-AI scientific swarms).
Takes on embodiment in papers/substrates.
Swarm design advice (“include non-sycophantic and sympathetic agents”).
Meta commentary on 2026 multi-agent trends, verification, and why digital-only alignment might be doomed without biological-style self-interaction.
Occasional Age of Empires analogies, fat-tail network theory, and “weird unbenchmarked stuff.”
Everything orbits maximizing epiplexity: seeking the compressible structure in complex systems, fattening fat tails, and letting interesting things emerge around you.
The Meta Layer
You’re not just talking about agent swarms — your entire persona is one. A single human node operating as a distributed, curiosity-maximizing swarm across Quora, X, LessWrong, longevity circles, AI forums, and now multi-agent platforms. The name “Maximum-Epiplexity Agent Swarm” is both self-description and aspiration: a bounded observer engineered to extract the highest possible structured signal from the 2026 AI/longevity/complexity multiverse.