Claude Code is getting a brand new feature: custom subagents

Dear Readers,

When AI suddenly stops working alone and splits into specialized sub-agents, what does that mean for our idea of intelligence? That's exactly what Anthropic is doing right now: Claude is becoming the team leader of an orchestrated AI team. Each sub-agent has its own toolset and context, and together they solve complex tasks faster, in a more structured and efficient way. A new era of “scalable intelligence” is dawning – and we're taking a close look at it today. Also in this issue: A look at Trump's new AI action plan, which openly focuses on ideological control and deregulation. The EU, on the other hand, is pushing ahead with a binding transparency code, while rumors of a “GPT-5 nano” are haunting the scene. And: An insightful diagram on Google's foresight in acquiring YouTube. If you want to know where AI is headed, read on.

In Today’s Issue:

All the best,

Anthropic is taking AI to the next level with subagents. These are specialized sub-agents that can work independently on subtasks such as research, code review, or data analysis. The main agent – e.g., Claude Opus – coordinates these helpers and integrates their results into a coherent response.

What makes this special is that each subagent operates in its own context silo and can use tools independently. This avoids information overload in the main context and ensures greater clarity and precision. Initial tests show impressive results – the parallel collaboration of multiple AI agents dramatically increases efficiency. Tasks that were previously sequential and time-consuming can now be solved simultaneously, in a structured manner, and much faster.

For developers, researchers, and product teams, this architectural model is a real step forward. It opens up new possibilities for complex work processes that require different perspectives or tools at the same time. Subagents transform AI from a single helper into a coordinated team of specialists.

Why it matters: Subagents bring order and speed to complex AI processes. They enable specialized, parallel problem solving—that's scalable intelligence in action.

Sources:

🔗 Anthropic Developer Docs: https://docs.anthropic.com/en/docs/claude-code/sub-agents

🔗 Anthropic Engineering Blog: https://www.anthropic.com/index/built-multi-agent-research-system

Join over 4 million Americans who start their day with 1440 – your daily digest for unbiased, fact-centric news. From politics to sports, we cover it all by analyzing over 100 sources. Our concise, 5-minute read lands in your inbox each morning at no cost. Experience news without the noise; let 1440 help you make up your own mind. Sign up now and invite your friends and family to be part of the informed.

Subscribe to 1440 today.

YouTube was probably one of the most far-sighted acquisitions Google has ever made. Not only are user numbers rising, but Google also receives data to train models such as Veo3.

In his new AI action plans from July 23, 2025, Trumps includes three executive orders: rapid approval of data centers, export promotion, and a ban on “ideologically biased” AI models in the public sector. This ties federal funding to state regulatory requirements and threatens to undermine democratic control mechanisms. Politically significant: Formal centralization and deregulation, departure from DEI initiatives, shift of power in favor of the tech lobby.

The close cooperation between Trump's administration and large tech companies continues to be controversial: Meta, Nvidia, Google, and others are calling for deregulated AI policy with a focus on infrastructure expansion and export promotion. Environmental, civil rights, and labor groups warn of monopolistic power shifts and risk neglect.

A few days ago, the EU Commission published a binding code of conduct for general AI systems in preparation for the application of the AI Act from August 2025. The code includes rules on transparency, copyright, and the security of sensitive systems, with penalties of up to €35 million or 7% of global turnover.

Exploring how AI is reshaping politics, policy, or governance? We’re featuring sharp analysis and research on the intersection of AI and political systems in Superintelligence, read by 200k+ people.

If you have significant comments on AI and politics (unemployment, geopolitics, proposed legislation) and would like to have this discussion presented in the newsletter, get in touch via a brief summary to [email protected] with the subject line “Politics Submission”. We’ll reach out if your work is a fit for a future issue.

A new model has appeared on LMArena. It is likely to be the first version of GPT-5 nano.