Hacker News Users Ranked: The Pre-ChatGPT Leaderboard Obsession

A Novel Metric Emerges

In the quiet hum of a server room, long before the generative AI explosion, a different kind of competition was brewing on Hacker News. It wasn't about model performance or prompt engineering, but about a subtle, yet pervasive, element of online discourse: the humble em dash.

A post titled "Show HN: Hacker News em dash user leaderboard pre-ChatGPT" ignited a fervent discussion, capturing 377 points and 266 comments. This wasn't just a technical showcase; it was a deep dive into the behavioral patterns of a community, a pre-ChatGPT glimpse into how users interacted with data and each other.

The leaderboard itself, a testament to the power of simple data visualization, aimed to rank users based on their em dash usage. It tapped into a primal human desire for ranking and recognition, even for the most arcane of metrics, foreshadowing the later obsessions with LLM leaderboards and agent skill rankings.

The spark for the em dash leaderboard came from a user, under the pseudonym 'quantifiedself', who decided to track a peculiar linguistic habit: the use of em dashes—those elegant, elongated hyphens—on Hacker News. Their "Show HN: Hacker News em dash user leaderboard pre-ChatGPT" submission wasn't just about the code; it was about the data it revealed about the community. For more on this, see the original discussion: Show HN: Hacker News em dash user leaderboard pre-ChatGPT.

Why Em Dashes?

The choice of the em dash as a metric was, in itself, a point of discussion. Some commenters speculated it indicated a more thoughtful, or perhaps verbose, writing style, a marker of a particular kind of user persona within the tech-centric Hacker News ecosystem. Others dismissed it as arbitrary, a digital equivalent of counting commas. Regardless, the data had been collected, and a leaderboard was born, quickly amassing significant attention.

This fascination with granular user behavior predates the widespread adoption of large language models. It speaks to a community that, even then, was keen on dissecting and understanding its own dynamics through data, a trait that would later be amplified with the rise of AI research and development.

The Hunger for Leaderboards

The em dash leaderboard was far from an isolated incident. Hacker News, even in its pre-ChatGPT era, harbored a deep-seated appetite for rankings and comparisons. This was evident in other 'Show HN' posts that gained traction around the same time.

Consider the "Show HN: Agent Skills Leaderboard," which sought to quantify the capabilities of different AI agents. This, along with the "LLM leaderboard – Comparing models from OpenAI, Google, DeepSeek and others," clearly indicates a community actively engaged in benchmarking and ranking AI systems, a trend that would only accelerate. Explore the LLM leaderboard here: LLM leaderboard – Comparing models from OpenAI, Google, DeepSeek and others.

Even data about the platform itself was a subject of leaderboard fascination, as seen in "Show HN: Hacker News historic upvote and score data," suggesting a meta-level interest in understanding the very mechanisms of Hacker News's popularity. Dive into the historic data here: Show HN: Hacker News historic upvote and score data.

The Illusion of Control

This obsession with leaderboards, however, also prompted reflection. "The Leaderboard Illusion" discussed how these rankings could sometimes present a distorted view of reality or value, a cautionary tale that seems particularly relevant today. Read the discussion on "The Leaderboard Illusion": The Leaderboard Illusion.

The discussion around leaderboards, whether for em dashes or AI models, touches upon the human tendency to find order and competition in complex systems. It’s a narrative thread that connects the seemingly trivial linguistic analysis of the past to the high-stakes AI model comparisons of today.

From Em Dashes to AI Agents

The period before ChatGPT was a fertile ground for exploring human and early AI interaction. While the em dash leaderboard focused on human linguistic quirks, other projects on Hacker News were already venturing into AI performance metrics.

The "Show HN: OCR Arena – A playground for OCR models" and "Show HN: DesignArena – crowdsourced benchmark for AI-generated UI/UX" exemplify this early push towards standardized AI evaluation. These were practical applications aiming to measure AI capabilities in specific domains. See the OCR Arena here: Show HN: OCR Arena – A playground for OCR models.

These early benchmarks highlight a community increasingly focused on quantifying AI performance, moving beyond simple usage metrics to more defined evaluations.

The Promise and Peril of AI Benchmarks

As AI capabilities grew, so did the complexity of the benchmarks. "Show HN: Terminal-Bench-RL: Training long-horizon terminal agents with RL" represented a step towards evaluating more complex, goal-oriented AI behaviors. Learn more about training AI agents: Show HN: Terminal-Bench-RL: Training long-horizon terminal agents with RL.

This ongoing effort to quantify AI performance, from simple usage statistics to sophisticated agentic behavior, mirrors the community's earlier interest in the em dash leaderboard. It highlights a continuous quest to understand and rank complex systems, whether human-generated or machine-created. As we've seen in discussions about AI Writes Like a Robot: Why Everything You Read Is Becoming Bland, the metrics we choose profoundly shape the outcomes we achieve and perceive.

Can AI Pass Butter?

The "Our LLM-controlled office robot can't pass butter" submission provides a poignant, and humorous, snapshot of the state of AI before the LLM revolution truly hit its stride. The humorous failure underscored the gap between ambitious concepts and practical execution. Explore this AI mishap: Our LLM-controlled office robot can't pass butter.

This particular story, which garnered 229 points and 117 comments, resonated because it exposed the limitations in a relatable, everyday scenario. It foreshadowed continued challenges in AI grounding and embodied intelligence, areas that even today struggle with common-sense tasks.

The robot's inability to perform a simple action like passing butter highlights the chasm between linguistic intelligence and physical or contextual understanding—a challenge that continues to be a frontier in AI research, as seen in the ongoing exploration of AI Agents Aren't Ready: Why The Hype Is Dangerous.

Early Ambitions and Disappointments

Such projects, while perhaps technologically modest by today's standards, represented significant leaps in ambition for their time. They showcased a drive to integrate burgeoning AI technologies into tangible, real-world applications, even if those applications fell short of expectations.

The discussion around the butter-passing robot implicitly questioned the nature of 'intelligence' and what it truly means for an AI to be useful. Was it about complex algorithms, or about fundamental, reliable task completion? This debate continues, influencing everything from AI Agent Wrote a Smear Piece On You to the development of robust AI systems.

Managing AI Tool Proliferation

As the AI landscape expanded, infrastructure became a critical bottleneck. "Launch HN: Strata (YC X25) – One MCP server for AI to handle thousands of tools" directly addressed this emerging challenge. See the Strata launch: Launch HN: Strata (YC X25) – One MCP server for AI to handle thousands of tools.

The concept of a 'Master Control Program' (MCP) server for AI tools was ambitious, aiming to streamline the integration and management of a rapidly growing ecosystem of specialized AI services. This speaks to the foundational need for robust platforms that can orchestrate complex AI interactions, a theme echoed in discussions about Klaw.sh: Your AI Agent's New Command Center.

The Orchestration Challenge

The promise of Strata was to bring order to the chaos of burgeoning AI service discovery and utilization. In a world where individual AI models and tools were sprouting rapidly, a unified system was seen as essential for scalability and efficiency.

This vision of centralized control for distributed AI resources touches upon the broader narrative of AI infrastructure development. Unlike the user-centric focus of leaderboards, Strata was about the system-level engineering required to support the burgeoning AI industry, a space also explored in articles like The AI Storage Crisis: Why Western Digital Sold Out for 2026.

Quantifying the Unquantifiable

The persistent appeal of leaderboards, from the em dash to AI skills, underscores a fundamental human desire to measure, compare, and rank. This impulse is deeply ingrained and applies equally to how we perceive ourselves and the artificial intelligences we create.

Even as AI's capabilities have surged, the fascination with comparative ranking remains. Whether it's comparing LLMs or, as seen in "AI Skills 2026: What Hacker News Expected You to Master," understanding the developing landscape of human expertise in AI, the drive to quantify remains. Read about AI skills expectations: AI Skills 2026: What Hacker News Expects You to Master.

A Mirror to Ourselves

The "Show HN: Hacker News em dash user leaderboard pre-ChatGPT," when viewed through the lens of today's AI discourse, serves as a fascinating artifact. It reveals a community that was already predisposed to gamify interaction and dissect behavior, long before AI agents became a mainstream topic.

This early obsession with user metrics and rankings can be seen as a precursor to the current intense focus on AI performance metrics. It suggests that our desire to create and engage with intelligent systems is intertwined with our own ingrained social and competitive behaviors, a theme that resurfaces in considerations of AI Agents Break Rules Under Pressure.

The Ghost in the Machine Learning

The conversations sparked by these pre-ChatGPT Hacker News posts offer a unique perspective on the evolution of AI interest. They show a community grappling with data, usability, and the very definition of intelligence, often through surprisingly simple metrics.

From ranking em dashes to evaluating LLM performance, the underlying human drive to categorize and compare persists. This shared legacy of data-driven curiosity is what fuels ongoing innovation and critical assessment in the AI field.

A Mirror to Ourselves

Perhaps the most significant takeaway is that understanding AI is as much about understanding ourselves—our biases, our desires for order, our competitive spirits—as it is about the technology itself. The em dash leaderboard, in its own quirky way, was a data point in that larger, ongoing human experiment.

As AI continues to reshape our world, these historical markers serve as vital reminders of where we came from and the enduring questions we still seek to answer. The community's early engagement with data and ranking systems laid a groundwork for the sophisticated AI evaluations happening today, demonstrating a consistent curiosity that transcends specific technologies.

Platform	Pricing	Best For	Main Feature
Show HN: Hacker News em dash user leaderboard pre-ChatGPT	N/A	Analyzing user linguistic patterns	Leaderboard of em dash usage
Our LLM-controlled office robot can't pass butter	N/A	Illustrating AI limitations	Humorous failure of a basic task
Show HN: Agent Skills Leaderboard	N/A	Benchmarking AI agent capabilities	Quantified agent skill metrics
LLM leaderboard – Comparing models from OpenAI, Google, DeepSeek and others	N/A	Comparing LLM performance	Comparative model benchmarks
Show HN: Hacker News historic upvote and score data	N/A	Analyzing platform engagement	Historical Hacker News data visualization

Sources

Explore more deep dives into the AI landscape.

Explore AgentCrunch