The key to AI's economic disruption is its "task horizon"—how long an agent can work autonomously before failing. This metric is reportedly doubling every 4-7 months. As the horizon extends from minutes (code completion) to hours (module refactoring) and eventually days (full audits), AI agents unlock progressively larger portions of the information work economy.

AI agents excel not because they are inherently more intelligent, but because they can exhaustively test possibilities without the cognitive fatigue that limits human performance. This 'relentless tedium' is a superpower for tasks like finding obscure bugs.

Anyone can build a simple "hackathon version" of an AI agent. The real, defensible moat comes from the painstaking engineering work to make the agent reliable enough for mission-critical enterprise use cases. This "schlep" of nailing the edge cases is a barrier that many, including big labs, are unmotivated to cross.

Once an AI agent is well-trained, the problem isn't a lack of ideas, but a relentless flood of high-quality ones. This creates a human bottleneck where the primary job shifts from ideation to curation and execution. The team can't keep up with the agent's productive output.

Long-horizon agents are not yet reliable enough for full autonomy. Their most effective current use cases involve generating a "first draft" of a complex work product, like a code pull request or a financial report. This leverages their ability to perform extensive work while keeping a human in the loop for final validation and quality control.

The most underappreciated AI breakthrough is the ability for an agent to autonomously launch and manage subordinate agents. This allows for complex, parallel task execution and quality checking without human intervention, removing the human-in-the-loop as a primary bottleneck and enabling exponential productivity gains.

While intricate software "scaffolding" can boost an AI agent's performance, progress is overwhelmingly driven by the core model. A new model generation typically achieves the same capabilities with simple prompts that previously required complex engineering.

While AI models excel at gathering and synthesizing information ('knowing'), they are not yet reliable at executing actions in the real world ('doing'). True agentic systems require bridging this gap by adding crucial layers of validation and human intervention to ensure tasks are performed correctly and safely.

Early agent attempts failed because their reliability was too low. Without a baseline of success ('escape velocity'), users won't try meaningful tasks, which starves the model of the crucial usage data and feedback needed for it to learn and improve.