Or: How I learned to stop worrying about my browser history and start asking it questions

Every morning at 6:47 AM, my phone buzzes. I silence it. At 6:52, it buzzes again. I check my email. By 7:15, I’ve scrolled through WhatsApp, glanced at the news, and searched for “best coffee beans Singapore” for the third time this week. By noon, I’ve opened seventeen browser tabs, sent forty-two messages, and generated roughly 3,000 data points about myself—most of which I’ll never see again.

This is what researchers call digital exhaust—the trail of data we leave behind simply by existing in the digital world. Like the smoke from a car’s tailpipe, it’s created as a byproduct of our online activities, behaviors, and transactions.

But here’s a thought that haunts me: What if that exhaust isn’t waste? What if it’s fuel?

I. The Overlooked Goldmine

Three months ago, I downloaded my complete ChatGPT history. All of it. Every question I’d ever asked, every conversation I’d had with the AI over eighteen months. The file was massive—hundreds of thousands of words documenting my curiosities, my struggles, my 3 AM coding questions, my half-baked ideas about data visualization.

I stared at that file and realized: I have no idea what was in there.

This is the paradox of the quantified self movement. We generate more personal data than ever before—wearable fitness trackers, sleep monitors, location data, and countless other sources have made self-tracking accessible to most people—but we have no systematic way to understand it.

Personal informatics research tells us people excel at collection but struggle with integration and reflection. We count our steps but don’t know why we feel tired. We log our meals but can’t spot the patterns that matter.

But what if AI agents could read our entire digital history in seconds and tell us stories about ourselves?

II. Vibe Coding for Analysis

This is where the workshop begins. Not with data science or statistics or Python libraries—though those will appear—but with a fundamentally different question: What happens when we stop treating code as an artifact we must carefully craft, and instead treat it as something we can conjure on demand?

We’ll be “vibe coding,” powered by the new generation of AI coding agents like Claude Code, Codex, and others that have transformed what’s possible. These agentic systems capable of reading entire codebases, understanding dependencies between files, and executing multi-step plans.

Here’s what this means in practice: Instead of spending hours writing data processing scripts, you describe what you want in plain language. The AI generates the code, tests it, fixes the bugs, creates visualizations, and even writes the narrative explaining what it found—all while you watch and guide.

The code becomes a byproduct, not the product. Analyzing your digital exhaust transforms from a technical challenge into a creative exploration.

III. What You’ll Actually Do

First, you’ll extract your digital traces. Your ChatGPT conversations, Google search history, browser data, WhatsApp messages, GitHub commits, or even your command-line history—any digital exhaust you’re comfortable examining. (We provide samples if you prefer.)

Second, you’ll talk to an AI about it. Not in code—in questions. “What patterns emerge in my search behavior over time?” “When am I most creative based on my commit messages?” “What topics dominate my conversations?” The AI generates code to answer these questions, creates visualizations, and tests hypotheses—all driven by natural language.

Third, you’ll discover stories. This is where it gets interesting. You’ll use prompts designed to uncover non-obvious insights—the kinds of patterns that make you lean forward and say “wait, really?” The workshop includes a comprehensive prompts (modeled after investigative journalism techniques) that guides the AI to:

Hunt for extreme distributions and surprising correlations
Find leverage points—small changes that could yield big effects
Stress-test findings against statistical fallacies and data biases
Package insights as narrative-driven data stories

Finally, you’ll compare and share. What patterns do you share with peers? What makes you unique? This optional extension transforms personal data analysis into collaborative discovery.

IV. Who Should Come (And Why)

This workshop is for three kinds of people:

Data enthusiasts who are tired of conventional sources. If you’ve analyzed every public dataset on Kaggle but never thought to analyze your own life, this is your invitation. Your digital exhaust is the most unconventional data you have access to—and it’s deeply personal.

AI practitioners curious about the full data value chain. You know LLMs can write code and generate text, but have you where they succeed and where they break in orchestrating an entire analysis pipeline—data prep, statistical testing, visualization, and narrative generation—while you provide high-level guidance? The workshop let you discover how LLMs can be creative collaborators across every stage of data work.

Educators, designers, and researchers working on personal informatics, digital well-being, or data ethics. If you’re thinking about how people relate to their own data, this workshop offers both practical techniques and philosophical questions. When AI can generate insights about us from our digital traces, what does self-knowledge mean?

V. The Prerequisites (Spoiler: Fewer Than You Think)

Here’s what you need:

A coding agent that runs in VS Code or your command line with sufficient credits. Options include Codex, Claude Code, GitHub Copilot, Google Antigravity, Google Jules, Cursor, etc. These tools have become remarkably accessible—many offer free tiers or trial periods.

Curiosity and openness to experimentation. This is listed as the only true prerequisite because it’s true. Programming knowledge helps but isn’t required. The whole point is that the AI handles the technical details while you focus on asking good questions.

Willingness to confront your digital self. Your digital exhaust will reveal patterns about you—some expected, some surprising, some possibly uncomfortable. The workshop creates a safe space for this exploration, but the most valuable insights often come from being willing to look closely at yourself.

VII. What You’ll Walk Away With

By the end of this workshop, you’ll have:

Hands-on experience with “code on demand”—a fundamentally different way of working with data where you describe what you want and AI generates the implementation.

A framework for analyzing personal data that balances creativity with rigor, using investigative journalism techniques adapted for data work.

Actual insights about yourself (or sample data) that you discovered through the process—stories hidden in your digital exhaust that you didn’t know existed.

And perhaps most importantly, you’ll understand viscerally what it means when code becomes a byproduct rather than an artifact—when the creative work shifts from implementation to interrogation, from syntax to sense-making.

At 6:47 AM tomorrow, my phone will buzz again. I’ll silence it. At 6:52, I’ll check my email. And this time, I’ll know that every search, every message, every browser tab is adding to a story I can now read.

The question isn’t whether you’re creating digital exhaust. You are. The question is: What will you do with it?

Mining digital exhaust

{{ gettext('Draft') }}