PLUS: Claude for Main Street, Medicare's AI overhaul, and an AI power grab
Happy reading
A new open-source framework is giving AI agents the ability to teach themselves new skills and improve over time. The system, called Hermes Agent, runs entirely on local hardware, marking a major step forward for personalized AI.
This signals a shift from static, command-based tools to dynamic AI partners that truly evolve with their users. What new possibilities for secure and private automation are unlocked when a powerful, self-improving AI can operate completely offline?
In today’s Next in AI:
A new self-improving AI agent framework
Claude's new tools for small businesses
How Medicare's new model funds AI healthcare
The AI boom’s real-world power grab
The Self-Improving Agent
Next in AI: A popular new open-source framework called Hermes Agent is enabling AI agents to improve over time by writing and refining their own skills, all while running locally on personal hardware.
Explained:
Hermes stands out with its ability to self-evolving its skills, saving learnings from complex tasks or user feedback to continuously adapt and improve its performance.
Its design uses contained sub-agents for focused tasks, which improves reliability and allows it to run efficiently with smaller context windows on hardware like NVIDIA RTX and DGX Spark.
Built to be model-agnostic, the framework integrates out-of-the-box with popular runtimes like Ollama and LM Studio, providing a straightforward path to setting up a local agent.
Why It Matters:
This marks a significant step away from static, task-based agents toward dynamic AI partners that learn and grow alongside you. The ability to run a self-improving agent privately on your own machine unlocks new possibilities for personalized and secure automation.
Claude for Main Street
Next in AI: Anthropic is moving beyond the chat window with Claude for Small Business, a new offering that embeds AI-powered workflows directly into essential tools like QuickBooks, PayPal, and HubSpot. This initiative aims to help owners automate tasks and reclaim their time.
Explained:
The system works through Claude Cowork as a toggle install, connecting directly to the tools businesses already use and shipping with 15 ready-to-run agentic workflows.
To bridge the adoption gap, Anthropic partnered with PayPal on a free course teaching owners how to integrate AI responsibly into their operations, available on-demand today.
As a public benefit corporation, Anthropic is also supporting solopreneurs through initiatives like the Solopreneurship Accelerator Program to ensure the benefits of AI reach everyone.
Why It Matters: This launch signals a major shift from standalone AI chatbots to deeply integrated assistants that work inside the software businesses already rely on. By targeting the backbone of the U.S. economy, this practical application of AI could give small businesses a powerful new competitive edge.
Medicare's AI Overhaul
Next in AI: The U.S. Medicare program is launching a new payment model designed to reimburse healthcare providers for using AI. This move could unlock a massive new market for AI-driven health solutions.
Explained:
The program, called ACCESS, shifts reimbursement from paying for clinician time to rewarding measurable health outcomes. This structure rewards organizations for keeping patients healthy with preventative care rather than just treating them when they are sick.
This change directly enables companies like Pair Team, which uses a voice AI agent to monitor patients between visits, coordinate referrals, and provide companionship—actions that were previously difficult to bill for.
While promising, the model faces challenges. Similar CMS innovation programs have a history of increased federal spending instead of savings, and the low reimbursement rates require companies to lean heavily on AI to be profitable.
Why It Matters: By creating a direct payment pathway for AI-driven care, the federal government is building a powerful incentive for tech innovation in healthcare. This could significantly accelerate the development and adoption of AI tools that manage chronic disease at a national scale.
AI's Power Grab
Next in AI: The AI boom's insatiable appetite for energy is creating real-world consequences, as a Nevada utility plans to stop supplying power to nearly 50,000 Lake Tahoe residents to meet the demands of new data centers.
Explained:
NV Energy, which provides 75% of Lake Tahoe's electricity, is ending its wholesale contract in May 2027 to redirect capacity to a growing data-center corridor serving companies like Google, Apple, and Microsoft.
Projections show data centers in Northern Nevada could drive 5,900 megawatts of new energy demand by 2033, with their share of the state's electricity use potentially rising from 22% to 35% by 2030.
A solution is complex because Lake Tahoe's utility is regulated by California, but its grid is located entirely within Nevada's system, leaving residents caught in a jurisdictional maze with no single entity accountable.
Why It Matters: This situation is a powerful preview of the physical-world conflicts that AI's enormous energy needs will create. As tech companies pour hundreds of billions into AI infrastructure, communities everywhere may face difficult trade-offs between supporting technological growth and ensuring local resource stability.
AI Pulse
Amazon pivoted its AI shopping strategy, axing the Rufus chatbot and launching a new Alexa-powered agent integrated directly into search results.
Princeton mandated proctoring for all in-person exams for the first time in 133 years, citing the proliferation of AI and personal devices as a primary catalyst for the historic policy change.
Meta faced internal protests over a new program that tracks employee mouse movements and keystrokes to gather training data for AI agents, just days before a new round of mass layoffs.
Meta sparked user backlash after a test of its new AI assistant on Threads revealed that users were unable to block or remove the AI account from their feeds.