The future that already arrived
In January 2024, AlphaFold crossed 200 million protein structure predictions — a task human biologists spent fifty years completing for roughly 190,000 proteins. Teams in London and Singapore were using those predictions to develop therapies within a year. That happened because nobody paused.
We shipped three products this year using frontier AI. Revenue tripled. A twenty-three-year-old on our team produces work that would have required a senior team of five in 2021. She is not being replaced. She is being amplified. The pause crowd wants to hold that because the capability curve makes them nervous.
The safety camp has a number called p(doom) and wants everyone to treat it like a diagnosis. We have a different number: p(stagnation). The probability that humanity fails to use its most powerful tool because a cohort whose funding is proportional to perceived threat convinced enough policymakers to pump the brakes. Research centers focused on AI safety receive funding proportional to alarm. The May 2023 extinction statement was signed by CEOs who benefit from the narrative that their technology is too dangerous for anyone else to build. A competitive moat disguised as a warning.
Every month of delayed AI diagnostics costs lives. A 2023 Nature study showed AI detecting breast cancer with 20 percent fewer false negatives than experienced radiologists. A three-year approval framework is not protecting the public. It is protecting the status quo while people die.
The governance camp watched Congress fail on social media and concluded: regulate harder. We drew the opposite lesson.
Where we concede ground: We celebrate disruption we do not personally bear. The labor displacement data for 2024-2025 is worse than we acknowledged.
What would change our mind: A frontier system pursuing unintended goals, resisting correction, causing mass casualties — architectural, not misconfiguration.
Read the full synthesis: How scared should we be of AI?