New Smart AI Report
Editor: Alan
[New Smart AI Digest] The "gold content" of your job is being drained by AI. Anthropic's latest report reveals a counter-intuitive truth: The more complex the tasks measured by years of education, the faster the AI accelerates. Compared to being directly replaced, "deskilling" is even more terrifying—AI takes away the joy of thinking, leaving you only with chores. But the data also points to the only way out: understanding human-machine collaboration can increase the win rate by ten times. In this era of surplus computing power, this is a survival guide you must understand.
Anthropic just released the "Economic Index Report" on its official website yesterday.
The report focuses not only on what people are doing with AI, but more on to what extent AI is actually replacing human thinking.
This time they introduced a new dimension called "Economic Primitives," attempting to quantify task complexity, required education levels, and the degree of AI autonomy.
The future of the workplace reflected behind the data is much more complex than the simple theory of "unemployment" or "utopia."
Harder Tasks, Faster AI Work
In our traditional perception, machines are usually good at repetitive simple labor, while appearing clumsy in fields involving profound knowledge.
However, Anthropic's data presents the exact opposite conclusion: The more complex the task, the more amazing the "acceleration" brought by AI.
The report shows that for tasks that only require a high school education to understand, Claude can increase work speed by 9 times;
And once the task difficulty rises to the threshold requiring a university degree, this acceleration multiple soars directly to 12 times.
This means that white-collar elite work, which originally required humans to think hard for hours, is exactly the field where AI currently "harvests" the highest efficiency.
Even if we take into account the failure rate of AI occasionally hallucinating, the conclusion remains unchanged: The explosive efficiency gain from AI on complex tasks is enough to offset the repair costs caused by its errors.
This explains why current programmers and financial analysts rely more on Claude than data entry clerks—because in these high-intellectual-density fields, the leverage effect displayed by AI is the strongest.
19 Hours
The "New Moore's Law" of Human-AI Collaboration
The most shocking data in this report is none other than the test of AI "durability" (Task horizons, measured by 50% success rate).
Standard benchmarks like METR (Model Evaluation & Threat Research) believe that current top models (like Claude Sonnet 4.5) will see their success rate drop below 50% when dealing with tasks requiring 2 hours of human time.
However, in Anthropic's actual user data, this time limit has been significantly extended.
In commercial API scenarios, Claude can maintain a majority win rate in tasks involving 3.5 hours of work.
And in the Claude.ai conversation interface, this number is surprisingly pushed to 19 hours.
Why is there such a huge gap? The secret lies in the intervention of "humans."
Benchmark tests are AI facing the exam paper alone, while real-world users decompose a huge complex project into countless small steps and correct AI's course through continuous feedback loops.
This human-machine collaboration workflow pushes the upper limit of task duration (measured by 50% success rate) from 2 hours to about 19 hours, approaching 10 times.
This may be what the future of work looks like: It is not that AI completes everything independently, but that humans learn how to drive it to finish a marathon.
Folding on the World Map
Poor Learn Knowledge, Rich Engage in Production
If we raise our horizon to the global level, we will see a clear and slightly ironic "adoption curve."
In developed countries with higher GDP per capita, AI has been deeply embedded in productivity and personal life.
People use it to write code, make reports, and even plan travel itineraries.
But in countries with lower GDP per capita, Claude's main role is "teacher," with a large number of uses concentrated on coursework and educational tutoring.
Besides the wealth gap, this is also a reflection of a technological generation gap.
Anthropic mentioned that they are working with the Rwandan government to try to let people there move beyond the simple "learning" stage and enter a broader application layer.
Because without intervention, AI is likely to become a new barrier: people in wealthy regions use it to exponentially amplify output, while people in underdeveloped regions are still using it to make up for basic knowledge.
Workplace Anxiety: The Ghost of "Deskilling"
The most controversial and most vigilant part of the report is the discussion on "Deskilling."
Data shows that the tasks currently covered by Claude require an average of 14.4 years of educational background (equivalent to an associate degree), which is far higher than the 13.2 years required for overall economic activity.
AI is systematically removing the "high-intellect" parts of work.
For technical writers or travel agents, this could be catastrophic.
AI takes over the "brain-using" work of analyzing industry dynamics or planning complex itineraries, leaving humans possibly only with trivial tasks like sketching or collecting invoices.
Your job is still there, but the "gold content" of the work has been drained away.
Of course, there are also beneficiaries.
For example, property managers. When AI handles boring administrative tasks like bookkeeping and contract comparison, they can focus their energy on client negotiations and stakeholder management that require high emotional intelligence—this is actually a kind of "Upskilling."
Anthropic cautiously stated that this is just a deduction based on the current situation, not an inevitable prophecy.
But the alarm it sounds is real.
If your core competitiveness is merely processing complex information, then you are right in the center of the storm.
A Return to the "Golden Age" of Productivity?
Finally, let us return to the macro perspective.
Anthropic revised their prediction of US labor productivity.
After excluding possible errors and failures by AI, they expect AI to drive productivity growth by 1.0% to 1.2% annually over the next decade.
This looks like a third reduction from the previous optimistic estimate of 1.8%, but do not underestimate this 1 percentage point.
This is enough to bring the US productivity growth rate back to the level of the Internet boom in the late 1990s.
Moreover, this is only based on model capabilities in November 2025. With the entry of Claude Opus 4.5, and as "Augmented Mode" (i.e., people no longer try to throw all work to AI, but collaborate with AI more smartly) gradually dominates user behavior, this number still has huge upside potential.
Conclusion
Going through the entire report, the most touching part is not that AI has become so strong, but how fast humans are adapting.
We are undergoing a migration from "passive automation" to "active reinforcement."
In this transformation, AI is like a mirror. It takes over tasks that require high education but can be completed through logical deduction, thereby forcing us to find values that cannot be quantified by algorithms.
In this era of surplus computing power, the scarcest ability of humans is no longer finding answers, but defining problems.
References:
https://www.anthropic.com/research/economic-index-primitives
https://www.anthropic.com/research/anthropic-economic-index-january-2026-report