Creativity · Science · AI · Tech
Where rigorous thinking meets liquid imagination.
I'm Carey Chou. I work where creativity, design, science, and human-centered AI meet, turning rigorous ideas into things people can understand and feel, and keeping the human at the center of the machine.
Science & Writing
Essays on AI systems: how rules, ML, GenAI and agents fit together; how to know an answer is good enough to act on; digital twins of expertise. Authored as HTML here, published in full fidelity to Medium.
Browse essays →Creative & AI Art
A studio for AI-assisted visual thinking: the liquid glass + flow concept, generative gradient fields, and interactive infographics that turn ideas into things you can feel.
Enter the gallery →The Science half
The writing library
Oct 2026 · Decision models · Interactive
Decision Reasoning Models: Giving LLM Agents Precise Decisions
LLM agents read a decision's context well but don't answer the same way twice. A small language model, fine-tuned to read that text and return one calibrated answer, shown on three decisions: a shopper's order of preference, a Yes or No on a conference, and a travel budget.
Explore →
Sep 2026 · Personalization · Interactive
Personalized Memory: A Model on Top of the Model
Even a huge model is still one average over millions of shoppers. A small per-customer memory sits on top and carries each shopper's own baseline, floored so a quiet session never collapses back to the average. A State Space / Kalman view of personalization, with no retraining.
Explore →
Sep 2026 · Data journalism · Interactive
Protein Mania: What 3,578 Labels Actually Say
Two million USDA branded-food records and three companies' audited filings. Only 36% of products sold on a protein claim clear FDA's own bar, and every figure carries a magic lens you drag to drill down in place.
Explore →
Sep 2026 · Human–AI collaboration · Interactive
Cognitive Orchestration: Knowing Which Decisions Are Settled
Agentic orchestration already answered how work moves between agents. Being cognitive is a different question: knowing which decisions are settled, which are genuinely contested, and which have no precedent at all. That gets measured per case at runtime, out of the disagreement every pipeline throws away as noise.
Explore →
Sep 2026 · Memory · Interactive
Collaborative Episodic Memory
We taught machines to reason with us, not to remember with us. What a session leaves behind turn by turn, and what happens when those episodes are finally read together.
Explore →
Sep 2026 · Signal study · Interactive
The GLP-1 Ripple: How One Class of Drug Is Re-pricing the American Plate
A molecule that edits appetite is, in economic terms, a demand shock with a prescription pad. A visual essay tracing the wave outward through the food industry, retail shelves, the job market and health & wellness.
Explore →
Sep 2026 · Reasoning · Interactive
Beyond Codifying Explicit Steps: Teaching AI the Implicit Decisions, Weights and Constraints
Most codification captures the steps experts can write down. This recovers the layer underneath: the criteria, conditional weights and constraints that actually decide.
Explore →
Aug 2026 · Evaluation
How Do You Know an AI Answer Is Good Enough to Act On?
An analyst asks an agent a question about a product category and gets back a clean answer: here are the strongest item affinities, here is what is different abo…
Read →
Jul 2026 · Framework · Interactive
The AI Cognitive Quadrant
A map for where rules, ML, GenAI, and agents actually fit, and what each one asks of the people around it.
Explore →
Jun 2026 · Digital Twins
From Digital Twin to Phygital Twin: Codifying Process Knowledge into Agentic Robotic Process…
RPA is brilliant once you already know the steps. Process mining is brilliant once the work already lives in clean system logs. Both quietly assume somebody wrote the process down, and in most companies nobody has. On capturing the scattered, tacit workflow that lives across a dozen apps and in people's heads, and turning it into an agent that can run it.
Read →
May 2026 · Digital Twins
From Super Teammate to MetaTwin: Toward Self-Building Digital Twins of Expertise
An open-source skill pack: withsuper-analyst, that could learn a dataset and produce a rules playbook. super-optimizer, which tunes a playbook against historica…
Read →
Feb 2026 · Adaptation
Test-Time Training: TTT-Discover
Most work with an LLM involves a frozen brain: weights fixed in time, and when it fails you prompt again and hope it gets lucky. TTT-Discover takes the other path and fine-tunes the model on the single problem in front of it, at inference.
Read →
Dec 2025 · Future of Work
The New Human Roles of AI: Why Boundaries, Not Models, Define the Future of Work
The Wall Street Journal recently published an article titled “Four New Jobs That May Be in Our AI Future.” https://www.wsj.com/tech/ai/new-jobs-ai-future-fefd4b…
Read →
Nov 2025 · Personalization
Nested Learning for Recommender Systems: Bringing Fast and Slow Learning to Personalization
Modern recommendation systems face a paradox:
Read →
Aug 2025 · Optimization
LLM-Driven Probabilistic Sampling for Human-Guided Optimization
Setting shelf levels for hundreds of products, you can't test every option. An LLM proposes candidates and a human steers the search. Probabilistic sampling for optimization that keeps a person in the loop.
Read →
Jul 2025 · Optimization
Evolving LLM Prompts to Generate Customer Shopping Narratives: AI-Guided Evolution with Cohesive…
In a previous article, we explored how combining large language models (LLMs) with reinforcement learning can create smarter business rules. Building on the Alp…
Read →
Jun 2024 · Experimentation · Interactive
Stop the Test When the Evidence Is In: SPRT and Mixture SPRT
Checking an A/B test every morning quietly turns a 5% false-alarm rate into 25%. Sequential testing fixes that, not by telling you to stop looking, but by changing what counts as enough.
Explore →The design lab
Interactions, played with first
The Magic Lens
Conventional drilldown answers “show me more” by replacing the page, taking the comparison that raised the question with it. A playable essay on drilling down in place. Drag a lens across five figures and watch detail arrive without the room going away.
Pick up the lens → Used for real in Protein Mania Liquid glassFlowInteractiveLiquid Glass & Flow
Five demos that take the site’s own design language apart: build the field, tune the glass over it one property at a time, freeze the motion to watch the effect fall over, and measure whether the text still passes contrast.
Move the sliders → Used for real in The GLP‑1 RippleThe blend
One mind, four instruments.
I don't think creativity and rigor are opposites; the best work needs both. Science gives ideas structure; creativity gives them a shape people want to touch; AI is the studio and the collaborator; tech is how it ships. This site is where those four meet.