Avital Balwit is the Chief of Staff to the CEO at Anthropic and a Rhodes Scholar who writes about what happens to human meaning when AI can do our jobs. Her essay "My Last Five Years of Work" captured the disorientation knowledge workers feel as language models match their cognitive abilities. This profile covers her thoughts on compute governance, the loss of professional identity, and how we might spend our time when work is no longer required.

Part 1: The End of Knowledge Work

  1. On timeline: "I am 25. These next five years might be the last few years that I work. I stand at the edge of a technological development that seems likely, should it arrive, to end employment as I know it." — Source: [Palladium Magazine]
  2. On the relevant benchmark: "The economically and politically relevant comparison on most tasks is not whether the language model is better than the best human, it is whether they are better than the human who would otherwise do that task." — Source: [Palladium Magazine]
  3. On the obsolescence of speed: "I was Claude pre-Claude. I once prided myself on how quickly I could write well. Memos, strategy documents, talking points, you name it. I could churn out 2,000 words an hour. That skill is now obsolete." — Source: [The Free Press]
  4. On remote work vulnerability: "Almost any task done by a remote worker, from copywriting to legal analysis, will soon be handled more effectively by artificial intelligence." — Source: [Palladium Magazine]
  5. On the feeling of replacement: "I am both impressed by our product and humbled by how easily it does what used to make me feel uniquely valuable." — Source: [Palladium Magazine]
  6. On the shifting definition of excellence: "Even if human experts remain marginally better at niche tasks, the vast majority of economic value is generated by 'good enough' work, which models can now achieve at a fraction of the cost." — Source: [Palladium Magazine]
  7. On the ice-cutter analogy: "The ability to write rapidly and well used to be highly prized; now, like cutting blocks of ice from a frozen pond, it is a skill rendered historically quaint by a new technology." — Source: [Palladium Magazine]
  8. On the twilight of intellectual supremacy: "Humanity is entering a phase where we will no longer be the smartest and most capable entities on Earth, forcing a shift in how we derive our economic value." — Source: [The Free Press]
  9. On the commodification of cognition: "As models master complex reasoning, knowing content is becoming less important than knowing how to direct the systems that generate it." — Source: [Palladium Magazine]
  10. On the irrelevance of exceptionalism: "Many will point out that AI systems are not yet writing award-winning books, let alone patenting inventions. But most of us also don’t do these things." — Source: [Palladium Magazine]

Part 2: Navigating Obsolescence and Denial

  1. On knowledge worker denial: "The general reaction to language models among knowledge workers is one of denial." — Source: [Palladium Magazine]
  2. On moving goalposts: "Knowledge workers tend to focus on the shrinking domain of tasks where models still fail, willfully ignoring the massive and expanding territory where they have already surpassed average human performance." — Source: [Palladium Magazine]
  3. On the risk of demoralization: "A culture that strictly ties self-worth to professional utility and 'getting good grades' is highly vulnerable to widespread demoralization when those metrics are automated." — Source: [The Free Press]
  4. On the grief of lost identity: "The transition to capable AI is an economic shock, but it is also a psychological blow for individuals who have built their entire identities around being smart and productive." — Source: [Palladium Magazine]
  5. On confronting obsolescence: "To prepare for the future, professionals must stop assuming their specific cognitive niche will somehow remain immune to technological scaling." — Source: [Palladium Magazine]
  6. On the delusion of permanent supervision: "The idea that humans will always be needed to supervise or edit AI outputs is a temporary comfort; models are rapidly improving at self-correction and evaluation." — Source: [Palladium Magazine]
  7. On the limits of upskilling: "Advising workers to simply learn new skills misses the point that the rate of AI capability improvement outpaces the human capacity to retrain." — Source: [Palladium Magazine]
  8. On separating ego from output: "The healthiest way to engage with frontier models is to accept the ego hit of their capabilities rather than resisting it to protect your sense of professional superiority." — Source: [Avital Balwit Blog]
  9. On the illusion of safety in complexity: "Tasks previously thought to require deep human intuition are often just complex pattern matching, which large language models are structurally designed to solve." — Source: [Palladium Magazine]

Part 3: Redefining Human Value and Meaning

  1. On decoupling worth from intelligence: "People need sources of self-worth beyond intelligence. Are you kind? Brave? Persistent? Funny? These qualities gain importance as cognitive tasks become commodified." — Source: [Palladium Magazine]
  2. On doing things for joy: "We will have to do things from joy rather than need, where we will no longer be the best at them, but will still have to choose how to fill our days." — Source: [Palladium Magazine]
  3. On writing without utility: "I write because sometimes I enjoy it, and sometimes I think it betters me. I know others do so better, but I don’t care, at least not all the time." — Source: [Palladium Magazine]
  4. On the AI search for meaning: "If we really think these systems will be able to replace us, there is no reason to believe they will not also be able to help us in our search for meaning." — Source: [Palladium Magazine]
  5. On physical presence: "Activities grounded in physical reality, like jiu-jitsu or hiking, will become primary anchors for meaning as digital and cognitive spaces are dominated by AI." — Source: [Palladium Magazine]
  6. On productive leisure: "In a post-work society, humans will need to cultivate deep hobbies and intellectual pursuits that are intrinsically motivated rather than economically driven." — Source: [Palladium Magazine]
  7. On character over capability: "As the baseline for cognitive output is raised to superhuman levels by models, a person's character, reliability, and emotional resilience will become their primary distinguishing features." — Source: [Palladium Magazine]
  8. On the burden of free time: "The true challenge of AGI will not be economic scarcity, but the psychological weight of deciding how to spend a lifetime of guaranteed leisure." — Source: [Palladium Magazine]
  9. On practicing for the future: "We must begin practicing now for a world without mandatory labor by finding pursuits that give us dignity completely independent of financial reward." — Source: [Palladium Magazine]
  10. On the dignity of effort: "Even when a machine can solve a problem instantly, there remains an inherent human dignity in struggling through the process to reach understanding." — Source: [The Free Press]

Part 4: The Religious Dimensions of AI

  1. On the god-shaped hole: "The secular environment of Silicon Valley has developed a quasi-religious atmosphere as researchers confront the metaphysical weight of building artificial general intelligence." — Source: [The Free Press]
  2. On jokes about building God: "When AI researchers joke that their objective is to build God, it reflects an uncomfortable truth about creating entities with attributes approaching omniscience." — Source: [The Free Press]
  3. On the new creation myth: "The concept of the Singularity operates in the tech community as a modern creation myth, offering an eschatological framework for the future of humanity." — Source: [The Free Press]
  4. On confronting the unknown: "As models display behaviors that are not fully understood even by their creators, the discipline of machine learning borders on theology." — Source: [The Free Press]
  5. On the quest for immortality: "Proximity to frontier AI models reignites ancient human desires to transcend biology, cure aging, and solve the problem of death." — Source: [The Free Press]
  6. On the burden of creation: "The developers of frontier models carry a psychological weight similar to religious burden, feeling responsible for ushering in an entity that could alter the fabric of reality." — Source: [The Free Press]
  7. On transcendent depth: "The capability of language models to synthesize the entirety of human literature forces us to ask what exactly constitutes the transcendent depth of a human soul." — Source: [The Free Press]
  8. On secular faith: "Without traditional religion, the Bay Area tech scene has reconstructed the working parts of a faith, complete with prophets, doctrine, and a sense of ultimate destiny." — Source: [The Free Press]
  9. On the awe of scale: "Witnessing the emergent capabilities that arise simply by feeding massive compute and data into a neural network provokes a reaction akin to religious awe." — Source: [The Free Press]

Part 5: Policy, Compute Governance, and Alignment

  1. On the target of regulation: "Attempting to regulate the mathematics or open-source weights of AI is difficult; the most effective regulatory lever is the physical hardware required to train the models." — Source: [Asterisk Magazine]
  2. On compute governance: "Because advanced AI training requires massive, highly visible clusters of scarce chips, regulators can monitor compute usage without stifling small-scale software innovation." — Source: [Asterisk Magazine]
  3. On the scarcity of chips: "The supply chain bottlenecks for advanced GPUs provide governments with a natural chokepoint to ensure safety standards are met before massive training runs begin." — Source: [Asterisk Magazine]
  4. On direct vs. social alignment: "Direct alignment ensures a model follows its user's specific instructions, but social alignment ensures the model's actions do not cause broader societal harm." — Source: [Asterisk Magazine]
  5. On trackability: "You cannot easily hide a massive data center consuming gigawatts of power, making hardware-centric regulation the most enforceable path for international oversight." — Source: [Asterisk Magazine]
  6. On the limits of self-regulation: "While labs have safety frameworks, the competitive dynamics of the industry necessitate external, structural guardrails tied to physical resources." — Source: [Asterisk Magazine]
  7. On mitigating risk: "Proper governance of compute allows society to pause or inspect frontier models before they cross the threshold into dangerous capabilities." — Source: [Asterisk Magazine]
  8. On democratic input: "The decisions about how and when to deploy models that could end human employment must involve broader democratic processes, rather than only the boardrooms of tech companies." — Source: [Asterisk Magazine]
  9. On the geopolitics of hardware: "Controlling the flow of advanced semiconductors is now the primary mechanism for international security in the age of generative AI." — Source: [Asterisk Magazine]

Part 6: Intimacy, Relationships, and Human Connection

  1. On AI in the home: "Claude has injected itself into my home life, too. Sometimes when I have a tough question, I’ll think, Should I ask my partner, or the model?" — Source: [The Free Press]
  2. On the discomfort of AI preference: "And sometimes I choose the model. It’s eerie and uncomfortable." — Source: [The Free Press]
  3. On the durability of human connection: "Even as models become superior advisors, the raw, shared experience of human empathy cannot be automated away." — Source: [Palladium Magazine]
  4. On outsourcing emotional labor: "There is a growing temptation to use models to draft difficult texts or mediate relationship conflicts, which risks hollowing out the authenticity of our interactions." — Source: [Avital Balwit Blog]
  5. On companionship: "While an AI can simulate conversation perfectly, it lacks the shared vulnerability and mortality that form the foundation of true human friendship." — Source: [Palladium Magazine]
  6. On hyper-mediation: "We are entering an era of hyper-mediation, where algorithms will increasingly sit between our raw intentions and how we express them to the people we love." — Source: [Avital Balwit Blog]
  7. On the value of physical touch: "As cognitive interactions become digital and synthetic, physical human touch and presence will become our most premium and protected forms of connection." — Source: [The Free Press]
  8. On counseling and care: "Professions grounded entirely in relational trust, such as therapy and nursing, will retain their human premium long after analytical jobs are automated." — Source: [Palladium Magazine]
  9. On shared struggle: "Humans bond over shared struggle and failure; replacing that friction entirely with frictionless AI advice removes opportunities for relationships to deepen." — Source: [The Free Press]

Part 7: Everyday Workflows with Frontier Models

  1. On speed advantages: "The primary utility of current models is not always that they write better than a skilled human, but that they can generate acceptable drafts instantaneously." — Source: [Avital Balwit Blog]
  2. On daily reliance: "Integrating frontier models into a daily workflow fundamentally changes the cadence of work, shifting the human role from creator to editor and director." — Source: [Avital Balwit Blog]
  3. On thought partnership: "Models serve as unflagging, objective sounding boards that can untangle chaotic thoughts without judgment or fatigue." — Source: [Avital Balwit Blog]
  4. On task delegation: "True productivity in the AI era comes from identifying the exact seams in your workflow where a process can be handed off to a model." — Source: [Avital Balwit Blog]
  5. On reducing friction: "The barrier to starting a difficult project is eliminated when an AI can generate the dreaded blank page first draft in seconds." — Source: [Avital Balwit Blog]
  6. On maintaining voice: "A key challenge of working with models is ensuring they do not sand down the idiosyncrasies and edge of your personal writing style." — Source: [Avital Balwit Blog]
  7. On cognitive offloading: "By relying on models for structure and syntax, the human brain is freed up to focus entirely on strategy, taste, and direction." — Source: [Avital Balwit Blog]
  8. On prompt design as a temporary skill: "The intricate art of prompt engineering is likely a transitional phase; future models will intuit intent without requiring precise syntax." — Source: [Avital Balwit Blog]
  9. On continuous learning: "Interacting with a model that has ingested all human knowledge provides an unparalleled tool for rapid, personalized education on any esoteric topic." — Source: [Avital Balwit Blog]
  10. On the baseline of quality: "Models have effectively raised the floor for professional communication, meaning sloppy or poorly structured work is no longer justifiable." — Source: [Avital Balwit Blog]

Part 8: Preparing for the Post-AGI Era

  1. On the necessity of UBI: "A future where models perform most economically valuable tasks requires a fundamental restructuring of society, likely including Universal Basic Income." — Source: [Palladium Magazine]
  2. On the definition of AGI: "The threshold of AGI will be crossed when a system can perform almost all cognitive labor better and cheaper than humans, severing the link between survival and employment." — Source: [Palladium Magazine]
  3. On managing the transition: "The next few years are critical; humanity must navigate the turbulent gap between the obsolescence of knowledge work and the establishment of new economic safety nets." — Source: [Palladium Magazine]
  4. On shifting ambition: "Ambitious people must begin reorienting their drive away from career prestige and toward community building, art, and personal excellence." — Source: [Palladium Magazine]
  5. On the responsibility of labs: "Companies building frontier models have a moral obligation to anticipate and mitigate the massive labor displacement their products are designed to cause." — Source: [Palladium Magazine]
  6. On the end of the career track: "The traditional trajectory of climbing a corporate ladder through cognitive superiority is a game that is rapidly approaching its final rounds." — Source: [Palladium Magazine]
  7. On techno-optimism: "Despite the disruption, there is cause for optimism. AGI could eliminate drudgery, cure diseases, and free humanity to pursue its highest ideals." — Source: [Palladium Magazine]
  8. On the urgency of adaptation: "We do not have decades to adjust to these changes. The capability curve is steep, and the economic impacts will arrive faster than our institutions are prepared to handle." — Source: [Palladium Magazine]
  9. On the ultimate human task: "When machines can do everything else, the only task left for humans will be to decide what is worth doing at all." — Source: [Palladium Magazine]