Six years ago, in February 2020, the world was running normally. Markets were strong, offices were open, children were in school and people were planning trips. Anyone who said that everything would change within weeks would have sounded alarmist. Three weeks later, Covid had redrawn the map. That comparison is at the center of a widely shared essay published on February 9 by AI entrepreneur Matt Shumer, titled “Something Big Is Happening,” which argues that artificial intelligence is at a similar point: dismissed by many as overhyped while, for people working inside the industry, the change has already arrived.
Seen from that angle, the question is no longer whether AI is coming but how prepared anyone is for the way it is changing work.
From helping with work to doing it
Shumer’s central observation is that many practitioners have stopped making predictions and started describing their daily experience. He writes that for some technical tasks a person now states the desired outcome in plain language and the system produces a complete solution, rather than a draft that needs constant correction. If that holds, AI is moving from “helping with work” to “completing work,” and the shift is less about speed than about roles and responsibility. The human job moves from writing or building toward setting direction, judging quality, understanding risk and making decisions.
The objections
The essay has drawn criticism, and some of it is fair. Pandemic analogies carry emotional weight, and not every change arrives that way. Several commentators have pointed out that an epidemic is bounded by mathematics: exponential growth in infections eventually runs out of people to infect. Technology does not necessarily follow the same curve, in either direction. The comparison is a rhetorical device, not evidence.
A more durable point survives the critique. People judge new capabilities by their old experience. Someone who tried an AI tool in 2023 or 2024 and found it unreliable may have been right at the time, but a gap of a few months now represents a large difference in capability. The problem is not that earlier models made mistakes; it is that many people apply an old verdict to current tools without updating it.
Progress beyond coding
Shumer also argues that improvements do not stay confined to programming. As models improve steadily in one domain, the gains tend to transfer to others. Supporters cite measurements, such as the task-length charts published by the research group METR, suggesting that the length of work models can complete on their own has grown from the equivalent of minutes of human time to hours. Such estimates depend heavily on how the tasks are measured, and they should be read with that caveat. The underlying message is nonetheless consistent: the pace is accelerating rather than slowing, which is why the people getting real work out of these tools every day tend to be either the most confident or the most worried.
What a non-programmer should do
The most useful part of the debate is practical. Treat AI not as a substitute for search but as a working assistant that takes on whole tasks. A lawyer can ask for a draft and a counterproposal, not just a summary, and then verify them. An accountant can hand over a case file and ask for anomalies and risk points, then check the findings under their own professional responsibility. A marketer can request a full campaign strategy, content, scripts and an A/B testing plan, then refine it with human judgment. The gain comes from moving from “one question, one answer” to “one problem, one solution.”
It is also worth knowing that free tiers and default settings often do not show the level at which current models operate. Before reaching a conclusion, test the tools with a modern workflow and a capable model.
The point is preparation, not alarm. Whether the change becomes obvious in two years or ten, a sound strategy is to spend time each day applying AI to real tasks rather than reading or arguing about it, to look for small improvement opportunities inside your own job or business, and to use the tools to capture them. A person who can show in a meeting that three days of work now takes one, at better quality, has already moved ahead. Those who choose not to look are raising their own risk of falling behind.
