A recent viral moment centered on Jacob Coxon — a former employee of OpenAI and Anthropic — has renewed attention on internal tensions in the AI industry. Coxon resigned and posted his first-ever tweet, which gathered over 700,000 likes and roughly 140 million views, accusing AI companies of gambling with human lives. His action crystallized a dilemma that, the author argues, many inside the field are quietly facing.
The piece frames AI workers into two camps. The first group consists of people who would rather be doing something else but remain because they believe their presence reduces existential risk; they work to prevent AI-driven catastrophe. The second group consists of people who are wholeheartedly committed to building superhuman AI because it is what they most want to do; they believe the technology will save or transform humanity.
Coxon’s case is notable because he had seldom spoken publicly about AI risk before. His resignation and viral message prompted wider discussion about those who not only move between AI firms but leave the industry entirely for moral reasons. The media tends to cover transfers between companies (for example, the author notes moves such as Ilya Sutskever and Mira Murati leaving OpenAI to start new ventures, Andrej Karpathy joining Anthropic, and Dylan Scandinaro making the opposite move), yet less attention falls on the narratives of apostasy driven by ethical alarm.
A central claim is this: when the ‘preventers’ leave, company balance shifts and the zealots gain momentum. But another paradox emerges: as the singularity draws near (understood as the point when a superhuman AI’s intentions and actions become opaque), even the most dedicated builders may ask why they should sacrifice their remaining time if the event will occur with or without their labour. If the threshold is months or weeks away, spending those final months on 100-hour weeks can seem irrational; many would rather return to family, hobbies, or simply human pleasures.
From these dynamics the author sketches a possible mechanism for stalled, oscillating progress. Large-scale departures of humans who maintain and supervise systems can halt development, since even recursive self-improvement depends on human-maintained physical and institutional infrastructure. Some people may later return when danger seems lower; development resumes, approaches a threshold, and some leave again. The result is a two-state oscillation: perpetually imminent but never fully realized progress toward the singularity.
The piece summons a C. S. Lewis–attributed idea (adapted from reflections on the atomic bomb): if destruction is coming, let it find us engaged in sensible human activity — praying, teaching, reading, listening to music — rather than huddled and terrified. This moral observation supports the wider point that many will choose present human ends over participation in an inevitable technical transition.
In conclusion, the author suggests that the singularity may exert its strongest influence precisely from a distance at which it can never quite arrive. The internal conflicts and waves of resignations within AI companies could therefore produce a long-term pattern of stop-and-start development, shaping the trajectory and social consequences of advanced AI more by oscillation than by uninterrupted acceleration.
Why this matters
- The Coxon episode highlights that industry-internal moral dynamics matter alongside external regulation.
- Individual choices by engineers and leaders — whether to stay, leave, or return — can materially affect how fast and how safely development proceeds.
- If progress is oscillatory rather than steadily accelerating, policymakers, institutions, and companies may need to rethink risk management strategies and timelines.
Overall, the article argues that internal departures and conflicting motives create a paradox: the nearer the singularity seems, the more it reshapes human behaviour in ways that may keep it perpetually at arm’s length.
