In 2020, there was a moment that became a kind of cultural shorthand. Professionals across every industry, working from home for the first time, accidentally exposed themselves on video calls — not because they were stupid, but because they had never needed to learn the technology before. Within months, knowing how to mute yourself, share your screen, and manage a video call became baseline professional competency.
That moment is happening again. The technology is different. The exposure is different. But the gap between people who have learned to use it and people who have not is widening in exactly the same way.
What Dario Amodei Said
The CEO of Anthropic, one of the most important AI companies in the world, has stated publicly that he expects AI tools to eliminate fifty percent of entry-level office jobs within one to five years. Not someday. Within this decade, likely within the first half of it.
That number is not a prediction from a futurist or a pundit. It is an operational forecast from someone who builds these systems for a living. And it is not being said to alarm people. It is being said because it is true, and because the people most at risk are the ones who have decided the tools are not for them.
What Is Already Happening
The capabilities that used to separate AI from human professionals are collapsing with each model generation.
- In 2022, these models failed at basic math and could not reliably reason through multi-step problems.
- In 2023, they were passing bar exams and producing publishable research summaries.
- In 2024, they were writing production-quality code, generating financial models, and explaining complex scientific topics at an expert level.
- In 2025 and 2026, they are completing autonomous tasks that previously required a trained professional — medical image analysis, legal document drafting, software architecture, customer service at scale.
Data from METR, an organization that measures AI capability, shows that the duration of tasks AI can complete autonomously is doubling approximately every four to seven months. That is not linear growth. That is the kind of trajectory that reshapes industries before people realize it is happening.
My House Is Already Living This
I have three daughters. Here is what their relationship with AI looks like right now, without me forcing any of it.
Samantha is thirteen. She operates an LLC for a baking business. She uses ChatGPT to develop recipes, analyze her pricing for profitability, and understand her margins. She uses Canva for logo design and Gemini for marketing copy. She is thirteen years old and she is running a business with better AI support than most small business owners have.
Savi is nine. She uses video tools and ChatGPT to plan content strategy for her YouTube channel. She thinks about what her audience wants, generates ideas, and iterates on them the way a content team would.
Scarlet is seven. She asks Alexa for custom AI-generated bedtime stories. She has never known a world where she could not ask a machine to create something for her on demand.
These are not extraordinary children. They are ordinary children who were not told these tools were too complicated or not for them. The founders who thrive in the next decade will be the people who approach AI the way my daughters do — as infrastructure, not novelty.
What to Actually Do
The recommendations here are not abstract.
For adults in the workforce: Invest in paid AI tools. Claude Opus and GPT-5 are not the same as the free tier. Apply them to your actual work immediately. Do not use them to write party invitations and call it AI adoption. Use them for the hardest, most valuable work you do, and figure out how they change what is possible.
For parents: Encourage your children to use these tools in schoolwork. Not to cheat — to build. The students who know how to direct, evaluate, and iterate with AI will have a different kind of advantage than students who learned to produce outputs without any assistance. One of those skill sets will be valued in the economy they enter. The other will not.
Financially: Build reserves. Do not assume current income is guaranteed by the job you currently have. The roles most at risk are the ones that involve predictable, repeatable cognitive tasks. If you are in one of those roles, diversify your skills and your income before you need to.
On career paths: The advice to follow genuine passion rather than economic safety is actually practical now in a way it has never been before. The traditional career paths that offered security through routine are the ones most exposed. The work that requires genuine creativity, judgment, and human connection is the work that remains.
The Zoom Comparison Holds
In 2020, the people who struggled most were not the least intelligent people in the room. They were the ones who had decided the technology was not important to learn before it became unavoidable.
The same pattern is playing out now, at a much larger scale and with much higher stakes. The gap between people who use these tools and people who do not is not a gap in intelligence. It is a gap in adaptation.
The tools are not going to wait for everyone to feel ready. They did not wait in 2020 either.