Patrick Wood's recent piece hits on a truth that should make every thinking person pause: much of today's AI is now being written by AI. We have entered the era of recursive self-improvement: systems designing better systems, code writing better code, intelligence bootstrapping itself at speeds humans can barely comprehend. This is not science fiction. It is happening now, and it marks a pivotal chapter in the story of human obsolescence.
The implications are profound. For the first time in history, humanity is actively building tools that can outperform us at the very task of creation, the one domain we long believed was uniquely ours. Once AI systems can improve their own architecture, debug their own flaws, and generate superior successors, the feedback loop becomes exponential. Moore's Law on steroids. The singularity isn't some distant theoretical event; it's the logical endpoint of recursive self-improvement.
This path leads straight to humans replacing themselves.
We already see the early signs. Entire industries: coding, writing, design, analysis, even basic research, are being automated at breathtaking speed. What begins as assistance quickly becomes replacement. Companies discover they can do more with fewer humans. Creative fields once thought safe are falling. The economic logic is merciless: if a machine can do the job better, faster, and cheaper, capital will choose the machine. The result is not just job displacement. It is a fundamental reordering of human purpose.
The Luddites warned us about this 200 years ago. Their solution, smashing the machines, was crude and ultimately futile. Today's neo-Luddites and anti-tech ranters fare little better. They rage against AI on social media (ironically powered by the very technology they decry), warning of lost jobs, lost meaning, and lost humanity. Some of their concerns are valid: mass technological unemployment, the erosion of human skill, the risk of AI misalignment, and the spiritual emptiness of a world where machines do all the thinking. But their rage is often performative and selective. They rarely offer realistic alternatives. Technology has always displaced old ways of life. The printing press destroyed scribes. The automobile killed the horse economy. Progress is disruptive by nature.
The real question is not whether we can stop recursive self-improvement. We can't, short of a global catastrophe or totalitarian ban that would only drive the technology underground. The real question is how we navigate it.
Humans have always replaced parts of themselves with tools. We outsourced memory to writing, calculation to computers, and physical labour to engines. AI is the next logical step: outsourcing cognition itself. The danger lies not in the technology, but in our lack of wisdom about what makes life worth living once the machines surpass us at nearly everything.
A wiser path forward rejects both naive techno-utopianism and hysterical Luddism. We should accelerate beneficial AI while building robust guardrails. We must redesign education, economics, and culture for a world of radical abundance and human obsolescence in traditional work. Most importantly, we need to rediscover what is irreducibly human: creativity driven by soul rather than algorithm, relationships, meaning-making, love, courage, and the pursuit of truth and beauty for their own sake.
Recursive self-improvement is coming whether we like it or not. The machines are learning to build better machines. The road of humans replacing themselves is already paved. The only real choice left is whether we walk that road as passive victims or as conscious architects of a future where humanity thrives alongside, or perhaps beyond AI.
https://patrickwood.substack.com/p/what-most-ai-is-now-written-by-ai