Tech developer Simon Willison (co-creator of Django) has demonstrated a "mildly dystopian" but effective method for auditing online personas. By feeding 1,000 recent Hacker News comments into Claude 4.6, Willison generated a comprehensive psychological and professional profile that captures everything from technical biases to personal habits. The experiment highlights the increasing capability of large language models to synthesize fragmented public data into cohesive, insightful identities.

The workflow utilizes the Algolia Hacker News API, which allows anyone to fetch a user’s entire comment history via simple JSON requests. When processed by a high-context LLM, these thousands of words reveal recurring themes like "agentic engineering"—a term Willison uses for AI-assisted coding—and specific security anxieties regarding prompt injection. The model even correctly identified Willison’s habit of coding on his iPhone while walking his dog, showcasing how modern AI can connect disparate dots across a timeline of activity.

While the results are useful for identifying "bad-faith" commenters, the ease of this profiling raises significant privacy concerns. Even without private data, the sheer volume of public discourse provides enough signal for AI to deduce real-world identities and behavioral patterns. As context windows expand and reasoning improves, the barrier between public anonymity and total transparency continues to erode, forcing a re-evaluation of how we interact in digital spaces.

Tech developer Simon Willison (co-creator of Django) has demonstrated a "mildly dystopian" but effective method for auditing online personas. By feeding 1,000 recent Hacker News comments into Claude 4.6, Willison generated a comprehensive psychological and professional profile that captures everything from technical biases to personal habits. The experiment highlights the increasing capability of large language models to synthesize fragmented public data into cohesive, insightful identities.

The workflow utilizes the Algolia Hacker News API, which allows anyone to fetch a user’s entire comment history via simple JSON requests. When processed by a high-context LLM, these thousands of words reveal recurring themes like "agentic engineering"—a term Willison uses for AI-assisted coding—and specific security anxieties regarding prompt injection. The model even correctly identified Willison’s habit of coding on his iPhone while walking his dog, showcasing how modern AI can connect disparate dots across a timeline of activity.

While the results are useful for identifying "bad-faith" commenters, the ease of this profiling raises significant privacy concerns. Even without private data, the sheer volume of public discourse provides enough signal for AI to deduce real-world identities and behavioral patterns. As context windows expand and reasoning improves, the barrier between public anonymity and total transparency continues to erode, forcing a re-evaluation of how we interact in digital spaces.