Lenin, perhaps apocryphally, observed that decades can pass with little change, while weeks can reshape the world. This sentiment perfectly captures the astonishing acceleration in generative AI over the past year – particularly the last few weeks – forcing even the most cautious observers to reconsider their expectations.
In September, Anthropic’s Claude, a leading AI model, demonstrated an unprecedented ability to sustain complex coding tasks for an astonishing 30 hours. Reports surfaced of the AI building entire web applications, generating over 11,000 lines of code in a single run. This capability reached a turning point in January when journalists, with no prior programming experience, successfully used Claude Code to build and launch a project for their publication, heralding a new era of “vibe coding” – software creation through simple description.
Simultaneously, OpenClaw gained viral attention as an open-source assistant seamlessly integrating with everyday communication platforms like WhatsApp and Slack to execute intricate, multi-step tasks. However, the most significant shift isn’t just in individual tools, but in the underlying architecture. The AI ecosystem is rapidly converging on open standards, paving the way for widespread integration.
One such standard, dubbed “MCP” – likened to the “USB-C of AI” – is now being downloaded nearly 100 million times monthly. This surge indicates a transition from exploratory AI projects to operational, real-world applications. Markets are keenly observing the emergence of AI agents as potential economic forces, reacting with both excitement and caution.
When Anthropic announced its expansion into lucrative sectors like finance, law, and life sciences, the market responded swiftly, wiping out $300 billion in value from software and data stocks. Economists and analysts, previously skeptical of an AI boom, are now revising their forecasts, acknowledging a potential turning point in technological advancement.
This rapid evolution is driving entrepreneurs and investors to explore opportunities, particularly within healthcare and life sciences, where they believe AI can address critical challenges. One startup’s approach to primary care offers a glimpse into a potentially transformative – or unsettling – future.
Primary care is facing a genuine crisis. Overworked and underpaid, physicians are leaving the profession in droves, seeking alternative practice models or abandoning clinical work altogether. Recruiting new doctors is becoming increasingly difficult each year.
Dr. Lisa Rosenbaum, in a powerful podcast series, eloquently captures the profound consequences of this decline. Her research demonstrates a direct correlation between losing a primary care physician and increased mortality rates, emergency room visits, and hospitalizations – highlighting the inherent health benefits of a sustained patient-doctor relationship. Alarmingly, most patients never establish a new relationship after losing their PCP.
But Rosenbaum’s deepest concern isn’t merely statistical. She identifies a crucial quality – the “good doctor” phenotype – not as a skillset, but as a fundamental approach to care. She describes a physician who takes comprehensive responsibility for a patient’s well-being, a doctor whose patients instinctively know they would want to be informed. For Rosenbaum, this intuitive connection is the essence of good medicine, and its absence threatens the very soul of the profession.
Her greatest fear is a future where the system devolves into an “artificial-intelligence-enhanced triage system devoid of a relational core.” This is precisely the vision Muthu Alagappan, co-founder of Counsel Health, is striving to deliver – albeit with the intention of improving patient access to care.
Alagappan points out that 100 million Americans currently lack a relationship with a doctor. Counsel Health utilizes AI to handle initial information gathering and clinical reasoning, functioning as an “extremely smart medical resident” that presents a plan for the doctor’s approval. The goal is to dramatically reduce the cost of primary care visits, potentially to less than a dollar per encounter, and enable doctors to see significantly more patients.
Alagappan believes that AI surpasses human cognitive capabilities in many aspects of primary care, reserving human intervention for physical tasks like vaccinations or wound care. He anticipates regulatory changes that will allow AI to assume even greater responsibility. In Utah, Doctronic is already pioneering this approach, using AI to prescribe renewals for 190 routine medications, backed by malpractice insurance and oversight safeguards.
As AI capabilities rapidly expand, the temptation to apply them wherever they are most efficient is strong. However, this approach risks subtly redefining professions based on tasks technology performs well. If AI excels at symptom processing, protocol matching, and prescription renewals, we might begin to define medicine solely by these functions – much like defining health by step counts and sleep scores.
We must be wary of letting the “affordances of the tools become the horizon of truth,” as Kate Crawford warns. This challenge extends to pharmaceutical research and development, where AI has shown promise in data-rich areas but struggles with scarce or conditional data.
It’s crucial to focus on what truly matters, rather than what technology readily delivers. One company, Unbound, based in the UK, is attempting to do just that. They’ve built a preventive health company around three core dimensions: physiology, agency, and connection.
Unbound distinguishes itself by measuring connectedness alongside traditional biomarkers, assessing social connection as a clinical input. Their medical director views health as an “emergent property of social integration, purpose, and communal regulation.” They’ve replaced the waiting room with a coffee shop and integrated community events into their core offerings, recognizing the importance of social environment.
Perhaps most notably, Unbound incorporates a “future self” exercise, a positive psychology intervention that encourages participants to envision their ideal future and identify barriers to achieving it. This process strengthens the psychological connection between present and future selves, enhancing goal clarity and motivation. They leverage technology – an app for tracking recommendations and integrating test results – but always with intentionality, guiding it towards a broader goal of holistic well-being.
While Unbound’s success remains to be seen, its approach resonates with a growing desire for something beyond relentless metric optimization. The principles – deepening connection, fostering agency, and attending to physiology – are universally applicable and could be adopted by established healthcare providers and digital platforms. Peloton, for example, possesses the community infrastructure but lacks a framework that extends beyond performance metrics towards genuine flourishing.
Generative AI is advancing at an unprecedented pace, forcing a recalibration of expectations. The opportunities are immense, particularly in addressing critical challenges like the primary care crisis. However, the central challenge lies in ensuring that technology serves human needs, rather than allowing those needs to be defined by technological capabilities. The risk of reducing health to what can be optimized is real, but a more complete and less reductive vision – one that prioritizes physiology, agency, and genuine human connection – is within reach.
