引入了“tokenmaxxing”这一核心概念，将AI生产力提升的本质定义为“最大化token消耗”。这打破了传统节省算力的思维，反直觉地认为用尽全力消耗token才能榨取AI的最大价值，本质上是在探讨如何将电力最高效地转化为智力劳动。

AI agent 完成任务的实际速度比人类快数倍——但这个事实几乎从未出现在主流 AI 能力讨论中。「2 小时时间地平线」被大众理解为「AI 能做人类 2 小时的工作」，但实际上 AI 可能只需 20-30 分钟就完成了这个任务。这意味着 AI 的实际生产力倍数远高于时间地平线数字所暗示的，而低估 AI 效率的讨论普遍存在。

Summary of "AI Fatigue is Real"

• The Productivity Paradox: AI significantly speeds up individual tasks (e.g., turning a 3-hour task into 45 minutes), but this doesn't lead to more free time. Instead, the baseline for "normal" output shifts, and the work expands to fill the new capacity, leading to a relentless pace.
• From Creator to Reviewer: Engineering work is shifting from "generative" (energizing, flow-state tasks) to "evaluative" (draining, decision-fatigue tasks). Developers now spend their days as "quality inspectors" on an unending assembly line of AI-generated code.
• The Cost of Nondeterminism: Engineers are trained for determinism (same input = same output). AI’s probabilistic nature creates a constant cognitive load because the output is always "suspect," requiring more rigorous review than code written by a trusted human colleague.
• Context-Switching Exhaustion: Because tasks are "faster," engineers now touch 6–8 different problems a day instead of focusing on one. The mental cost of switching contexts so frequently is "brutally expensive" for the human brain.
• Skill Atrophy: Much like GPS has weakened our innate sense of direction, over-reliance on AI coding tools can cause core technical reasoning and mental mapping of codebases to atrophy.
• Strategies for Sustainability:
  • Time-boxing: Setting strict timers for AI sessions to avoid "prompt spirals."
  • Separating Phases: Dedicating mornings to deep thinking and afternoons to AI-assisted execution.
  • Accepting "Good Enough": Setting the bar at 70% usable output and fixing the rest manually to reduce frustration.
  • Strategic Hype Management: Ignoring every new tool launch and focusing on mastering one primary assistant.

On ai, productivity and shorter work weeks and why that will not happen

• Task Expansion & Role Blurring: AI lowers the barrier to entry for complex tasks, leading employees to take on work outside their core expertise. Product managers and designers are now writing code, while researchers take on engineering tasks.
• Specialist Burden: This expansion creates a "cleanup" tax. For example, senior engineers now spend significant time reviewing, debugging, and mentoring colleagues who produce "vibe-coded" AI outputs, often through informal and unmanaged channels like Slack.
• The "Ambient Work" Phenomenon: Because AI interactions feel conversational and "easy," work has become ambient. Employees find themselves prompting AI during lunch, between meetings, or late at night, eliminating natural mental downtime.
• Intensified Multitasking: Workers are running multiple AI agents in parallel while simultaneously performing manual tasks. This creates a high sense of "momentum" but leads to extreme cognitive load and constant attention-switching.
• The Productivity Trap: AI acts as a "partner" that makes revived or deferred tasks feel doable. This creates a flywheel where people don't work less; they simply take on more volume, leading to "unsustainable intensity" that managers often mistake for genuine productivity.
• Sustainability Risks: The researchers warn that while AI feels like "play" initially, it eventually leads to cognitive fatigue, impaired decision-making, and burnout as the quiet increase in workload becomes overwhelming.

Hacker News Discussion

• Cognitive Fatigue: Users highlighted that "AI fatigue" is distinct from normal work tiredness. It stems from the "constant vigilance" required to audit AI output and the lack of a "flow state" due to unpredictable waiting times for generations.
• Executive Function Strain: Commenters noted that managing autonomous agents is more exhausting than manual work. One user compared it to Level 3 autonomous driving—you aren't driving, but you must remain "fully hands-on" to ensure the AI doesn't touch the wrong files or hallucinate.
• The Jevons Paradox: Several participants pointed out that as the "cost" of work decreases due to AI, the demand for work increases proportionally. Instead of saving time, workers are expected to triple their output, which leaves them more stressed than before.
• Management Expectations: A common theme was that leadership often mandates AI usage and pre-supposes productivity gains, leaving no room for cases where AI makes work slower or lower quality. This forces employees to "perform" productivity while working longer hours.
• Vibe Coding vs. Engineering: There is a heated debate between those who see "vibe coding" (prompt-heavy development) as a massive efficiency gain and veterans who argue it produces "average code" that becomes a maintenance nightmare in large, legacy codebases.

• The author identifies two primary personality traits: "The Builder" (focused on velocity, utility, and shipping) and "The Thinker" (needing deep, prolonged mental struggle).
• "Thinking hard" is defined as sitting with a difficult problem for days or weeks to find a creative solution without external help.
• In university, the author realized this ability to chew on complex physics problems was their "superpower," providing a level of confidence that they could solve anything given enough time.
• Software engineering was initially gratifying because it balanced both traits, but the rise of AI and "vibe coding" has tilted the scale heavily toward the Builder.
• While AI enables the creation of more complex software faster, the author feels they are no longer growing as an engineer because they are "starving the Thinker."
• The lack of struggle leads to a feeling of being stuck, as the dopamine of a successful deploy cannot replace the satisfaction of deep technical pondering.

Hacker News Discussion

• The loss of the "clayship" process: Commenters compared coding to working with clay; skipping the struggle means missing the intimacy with the material that reveals its limits and potential.
• The "Vending Machine" effect: Receiving a "baked and glazed" artifact from AI removes the human element of discovery and learning.
• Risk of mediocrity: There is concern that AI guides developers toward "average" or conventional solutions, making it harder to push for unique or innovative ideas without significant manual effort.
• The tradeoff of efficiency: While some view the current era as the best time for "Builders" who just want to see results, many veteran developers feel a profound sense of loss regarding the cognitive depth of the craft.
• Clear communication as a new skill: Some argue that interacting with AI requires a different kind of "thinking hard"—specifically, the need to express creative boundaries clearly so the model doesn't "correct" away the uniqueness of the project.