Anthropic has just released an "AI taking jobs report": the higher the education level, the more likely one is to be affected
TechFlow Selected TechFlow Selected
Anthropic has just released an "AI taking jobs report": the higher the education level, the more likely one is to be affected
In an era of surplus computing power, humanity's scarcest ability is no longer finding answers, but defining questions.
Navigating Web3 tides with focused insightsOriginal author: Xinzhiyuan
The 'value' of your job is being drained by AI. Anthropic's latest report reveals a counterintuitive truth: the more complex a task measured by years of education, the faster AI accelerates. More threatening than outright replacement is 'deskilling'—AI takes away the joy of thinking, leaving you only with menial tasks. But the data also points to the only way forward: mastering human-AI collaboration increases your odds of success tenfold. In this era of surplus computing power, this is an essential survival guide you must read.
Anthropic just released its "Economic Index Report" on its official website yesterday.
The report focuses not only on what people are using AI for, but more importantly, on how much AI is truly replacing human thought.
This time, they introduced a new dimension called 'Economic Primitives,' aiming to quantify task complexity, required education level, and AI autonomy.
The future of work revealed behind the data is far more complex than simple narratives of 'job loss' or 'utopia.'
The harder the task, the faster AI gets it done
In our conventional understanding, machines usually excel at repetitive, simple labor, while appearing clumsy in areas requiring advanced knowledge.
But Anthropic's data presents a completely opposite conclusion: the more complex the task, the more astonishing the 'acceleration' brought by AI.
The report shows that for tasks understandable with just a high school education, Claude can boost work speed by 9 times;
once task difficulty reaches the threshold requiring a college degree, this acceleration multiplier soars to 12 times.
This means that white-collar elite jobs, which previously required hours of human deliberation, are now the areas where AI achieves its highest 'harvesting' efficiency.
Even when factoring in AI's occasional hallucinations and failure rates, the conclusion remains unchanged: the massive efficiency gains AI brings to complex tasks are sufficient to offset the costs of fixing errors.
This explains why programmers and financial analysts today rely on Claude more than data entry clerks—because in these high-intellect-density fields, AI demonstrates the strongest leverage effect.
19 Hours: The New Moore's Law of Human-AI Collaboration
The most shocking data in this report comes from testing AI's 'endurance' (task duration, Task Horizons, measured at 50% success rate).
Typical benchmark tests like METR (Model Evaluation & Threat Research) suggest that current top models (such as Claude Sonnet 4.5) see their success rate drop below 50% when handling tasks taking humans two hours.
But in Anthropic's real user data, this time boundary is significantly extended.
In commercial API usage scenarios, Claude maintains over 50% success rate on tasks involving 3.5 hours of workload.
In the Claude.ai conversational interface, this number is astonishingly pushed up to 19 hours.
Why such a huge gap? The secret lies in human involvement.
Benchmark tests put AI alone facing an exam paper, whereas real users break down large, complex projects into countless small steps, constantly correcting AI's course through feedback loops.
This human-AI collaborative workflow extends the upper limit of task duration (measured at 50% success rate) from 2 hours to about 19 hours—nearly a tenfold increase.
This might be the true shape of future work: not AI completing everything independently, but humans learning how to steer it through a marathon.
Folded World Map: The Poor Learn Knowledge, the Rich Produce
If we zoom out to a global view, we see a clear and slightly ironic 'adoption curve.'
In developed countries with higher per capita GDP, AI has already deeply integrated into productivity and personal life.
People use it to write code, create reports, and even plan travel itineraries.
But in lower-income countries, Claude’s primary role is that of a 'teacher,' with most usage concentrated on homework and educational tutoring.
Beyond wealth disparities, this reflects a technological generational gap.
Anthropic mentions it is collaborating with the Rwandan government to help people there leapfrog the mere 'learning' phase into broader application layers.
Because without intervention,AI could become a new barrier: people in wealthy regions exponentially amplify output, while those in underdeveloped areas remain using it to catch up on basic knowledge.
Workplace Concerns: The Ghost of 'Deskilling'
The most controversial and cautionary part of the report is the discussion around 'deskilling.'
Data shows that tasks currently covered by Claude require on average 14.4 years of education (equivalent to an associate degree), significantly higher than the overall economic average of 13.2 years.
AI is systematically stripping away the 'high-intellect' components of jobs.
For technical writers or travel agents, this could be catastrophic.
AI takes over analyzing industry trends and planning complex itineraries—the 'brainwork'—leaving humans perhaps only with sketching drafts or collecting receipts.
Your job remains, but its 'value' has been hollowed out.
Of course, there are beneficiaries.
Take real estate managers, for example—when AI handles tedious administrative work like bookkeeping and contract comparison, they can focus their energy on high-emotional-intelligence activities like client negotiation and stakeholder management—a form of 'upskilling.'
Anthropic cautiously notes this is merely extrapolation based on current conditions, not an inevitable prediction.
But the warning bell is real.
If your core competency is merely processing complex information, then you're right at the eye of the storm.
A Return to the 'Golden Age' of Productivity?
Finally, let's return to the macro perspective.
Anthropic has revised its forecast for U.S. labor productivity growth.
After accounting for potential AI errors and failures, they project AI will drive annual productivity growth of 1.0% to 1.2% over the next decade.
This appears to be a one-third reduction from the earlier optimistic estimate of 1.8%, but don't underestimate this 1 percentage point.
It's enough to bring U.S. productivity growth back to the levels seen during the late 1990s internet boom.
And this is based solely on model capabilities as of November 2025. With the arrival of Claude Opus 4.5 and the growing dominance of 'enhanced modes'—where users no longer dump entire tasks onto AI but instead collaborate with it more intelligently—this figure has significant upside potential.
Conclusion
Reading through the entire report, what resonates most isn't just how powerful AI has become, but how quickly humans have adapted.
We are undergoing a migration from 'passive automation' to 'active augmentation.'
In this transformation, AI acts like a mirror—it takes over tasks that require advanced education yet can be completed through logical reasoning, thereby forcing us to seek values that algorithms cannot quantify.
In this era of surplus computing power, humanity's scarcest ability 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
Join TechFlow official community to stay tuned
Telegram:https://t.me/TechFlowDaily
X (Twitter):https://x.com/TechFlowPost
X (Twitter) EN:https://x.com/BlockFlow_News
