Anthropic launches research program to study AI's job impact after CEO predicts 50% of white-collar roles will vanish in 5 years. New data shows coding work already transforming as AI agents automate 79% of developer tasks.
Good Morning from San Francisco, Germany wants DeepSeek banned from app stores. The reason? Chinese data laws. Berlin's
Germany demands Apple and Google remove Chinese AI app DeepSeek from their stores, claiming it illegally ships user data to China. The move reflects growing global tensions over AI privacy and data sovereignty.
Anthropic launches research program to study AI's job impact after CEO predicts 50% of white-collar roles will vanish in 5 years. New data shows coding work already transforming as AI agents automate 79% of developer tasks.
đź’ˇ TL;DR - The 30 Seconds Version
🚨 Anthropic launches Economic Futures Program with $50,000 grants after CEO Dario Amodei predicted AI will eliminate 50% of entry-level white-collar jobs within five years.
đź“Š New research shows AI automation rates hit 79% in specialized coding agents versus 49% in general chatbots, analyzing 500,000 developer conversations.
đź’» JavaScript, HTML, and CSS dominate AI coding tasks, suggesting user-facing development jobs face earlier disruption than backend work.
🏢 Startups use advanced AI coding tools at 33% rate while enterprise adoption lags at 13%, creating potential competitive gaps.
🔬 Program will fund 20-50 global research grants and host policy forums in Washington and Europe this fall to develop economic response strategies.
⚡ AI's coding improvements could accelerate AI development itself, creating a feedback loop that speeds up broader economic disruption across industries.
Anthropic wants to study the economic wreckage that AI might cause. The company launched its Economic Futures Program on Friday, offering grants and hosting policy forums to research AI's impact on jobs and the economy.
The timing isn't coincidental. CEO Dario Amodei recently predicted that AI will eliminate 50% of entry-level white-collar jobs within five years. That forecast reportedly upset the Trump administration, which wants to encourage AI innovation to compete with China.
The program offers three approaches to understanding AI's economic effects. Research grants up to $50,000 will fund studies on AI's labor impact. Policy symposiums in Washington and Europe this fall will bring together researchers and policymakers to propose solutions. Strategic partnerships with research institutions will expand the network studying these questions.
New research from Anthropic shows software development is getting a preview of AI's economic impact. The company analyzed 500,000 coding conversations across its platforms and found patterns that suggest broader changes ahead.
Anthropic's coding agent, Claude Code, automates far more tasks than its general chatbot. About 79% of Claude Code conversations involved automation, where AI performs tasks directly. Only 49% of regular Claude conversations reached that level.
The difference matters because it shows what happens when AI tools become more specialized. As companies build more focused AI agents, they might automate more work rather than just assist humans.
Developers are using AI most for user-facing work. JavaScript, HTML, and CSS—languages that build what users see—dominated the coding conversations. Web development and user interface design were among the top tasks.
This pattern suggests jobs focused on building simple applications and interfaces face earlier disruption. If AI handles more front-end coding through "vibe coding"—where developers describe what they want in plain language—these roles might shift toward higher-level design work.
The research revealed a striking adoption gap. Startups accounted for 33% of Claude Code usage, compared to only 13% for enterprise work. Traditional companies are moving more cautiously, likely due to security reviews and approval processes.
This mirrors past technology shifts where nimble startups gained competitive advantages by adopting new tools first. But AI's general-purpose nature could make this gap more consequential. Companies that integrate AI agents successfully might pull far ahead of slower adopters.
Sarah Heck, Anthropic's head of policy programs, said the goal is getting more people to study AI's broad effects. The company wants to understand how quickly AI adoption is happening and whether labor markets are already changing.
Even when AI automates coding tasks, humans stay involved more than expected. About 36% of Claude Code interactions involved "feedback loops"—where AI completes work but humans review results and send back error messages or corrections.
This suggests that even advanced AI agents need human oversight. But it raises questions about how long this pattern will last. More capable AI systems will likely need less human input over time.
The distinction between automation and augmentation is blurring as AI agents become more sophisticated. Traditional automation replaced human work entirely. These new AI tools create hybrid workflows where humans and machines collaborate in complex ways.
Anthropic's research focused on coding because it's among the most developed uses of AI in the economy. But the lessons might preview changes in other occupations as AI capabilities expand.
Large tech companies now generate about 25% of their code using AI models. Some companies have started hiring fewer software developers because existing teams can accomplish more with AI assistance.
The Economic Futures Program will look at broader questions. Researchers might study AI's effect on GDP, adoption rates across industries, and labor market changes already underway.
Anthropic plans to award 20 to 50 research grants globally and provide free access to its AI models for analysis. The company will also host forums for researchers and policymakers to develop response strategies.
AI's impact on coding might create a feedback loop that speeds up AI development itself. Since AI research relies heavily on software, better AI-assisted coding could accelerate breakthroughs and create a cycle of faster progress.
This makes studying AI's economic effects more urgent. Changes that might have unfolded over decades could happen in years if AI capabilities compound quickly.
The program represents Anthropic's attempt to get ahead of these changes through research and policy development. But it also acknowledges that current approaches to understanding AI's economic impact aren't keeping pace with the technology's development.
Individual developers and students represent about half of the coding interactions in Anthropic's data. This suggests AI coding tools are spreading beyond just business contexts into personal projects and education.
The Economic Futures Program will need to address fundamental questions about work and economic structure. If AI can automate significant portions of white-collar work, society will need new frameworks for employment, income distribution, and economic opportunity.
Amodei's prediction about job losses reflects these broader concerns. The Trump administration's reaction suggests that AI's economic disruption is becoming a political issue alongside its technological and competitive dimensions.
The research grants will accept applications on a rolling basis, with initial awards in mid-August. Policy proposals for the fall symposiums are due July 25.
Anthropic is positioning itself to influence how society responds to AI's economic effects. But the company is also contributing to those effects through its AI models and coding tools.
Why this matters:
Q: How much money is Anthropic giving away in research grants?
A: Anthropic plans to award 20 to 50 research grants globally, each worth up to $50,000. Applications are accepted on a rolling basis, with the first awards going out in mid-August 2025.
Q: What exactly is "vibe coding" and why does it matter?
A: Vibe coding is when developers describe what they want in plain language and let AI handle the technical details. It's becoming common for user interface work and could eliminate jobs focused on simple app development.
Q: Why are startups adopting AI coding tools faster than big companies?
A: Startups can move quickly while established companies need security reviews and approval processes. The research shows startups account for 33% of advanced AI coding usage versus only 13% for enterprise work.
Q: What are "feedback loops" in AI coding and do they keep humans involved?
A: Feedback loops happen when AI writes code but humans review results and send back error messages or corrections. About 36% of advanced AI coding interactions work this way, but this might change as AI improves.
Q: When and where are the policy conferences happening?
A: Anthropic will host symposiums in Washington DC and Europe this fall. Policy proposals are due July 25, 2025. Selected authors will present their ideas for responding to AI's labor market effects.
Q: How did Anthropic analyze 500,000 coding conversations without violating privacy?
A: They used a privacy-preserving analysis tool that turns conversations into higher-level, anonymous insights. The tool identifies topics like "UI development" without revealing specific user details or code.
Q: Could AI coding improvements actually speed up AI development itself?
A: Yes, since AI research relies heavily on software development. Better AI-assisted coding could accelerate breakthroughs and create a feedback loop that speeds up AI progress across all industries.
New research reveals most people don't use AI for therapy—yet. Only 2.9% of Claude conversations involve emotional support, but the longest sessions hint at deeper connections ahead as AI capabilities grow.
MIT researchers monitored students' brains while they wrote essays with ChatGPT. The AI users showed weaker neural activity and couldn't quote their own work. When they switched back to writing alone, their brains stayed weakened.
Anthropic says multiple AI agents working together beat single models by 90%. The catch? They use 15x more computing power. This trade-off between performance and cost might reshape how we build AI systems for complex tasks.
AI models typically learn by memorizing patterns, then researchers bolt on reasoning as an afterthought. A new method called Reinforcement Pre-Training flips this approach—teaching models to think during basic training instead.
Get tomorrow's intel today. Join the clever kids' club. Free forever.
