Welcome back to Daily Zaps, your regularly-scheduled dose of AI news ⚡️ 

Here’s what we got for ya today:

• 🧞‍♂️ DOGE looking to use AI agents to replace federal workers

• 👩🏻‍💻 Wikipedia is using (some) generative AI now

• 💳 Visa wants to give AI agents your credit card

• 🤖 Amazon launches its most capable AI model

Let’s get right into it!

Sign Up | Advertise | Tools

GOVERNMENT

DOGE looking to use AI agents to replace federal workers

Anthony Jancso, a former Palantir employee and early recruiter for Elon Musk’s Department of Government Efficiency (DOGE), is now hiring technologists through his startup AccelerateX to develop AI agents capable of automating tasks currently handled by tens of thousands of federal workers. In a Palantir alumni Slack post, Jancso described the project as “DOGE orthogonal,” claiming it could free up 70,000 full-time employees for more impactful work by standardizing over 300 roles. AccelerateX, originally known as AccelerateSF and once supported by OpenAI and Anthropic for hackathons, has since rebranded with a focus on modernizing government systems and is reportedly partnering with Palantir.

Jancso and his cofounder Jordan Wick, another DOGE participant, have appeared at multiple federal agencies as DOGE pushes AI-driven reforms, including using chatbots, automating regulation rewrites, and even exploring tools for mass terminations. However, experts warn that deploying AI at this scale across diverse government agencies is both technically and procedurally complex. Critics argue AI agents still lack reliability for mission-critical tasks and could pose risks if adopted too aggressively. Jancso’s close ties to Musk, SpaceX, and Palantir—alongside Trump-aligned figures like Steve Davis—have drawn further scrutiny, especially given Palantir’s deep involvement in federal surveillance and data integration efforts. 

NON PROFIT

Wikipedia is using (some) generative AI now

The Wikimedia Foundation is integrating generative AI into Wikipedia’s editing process to support, not replace, its volunteer editors by easing tedious tasks like background research, translation, and onboarding. The goal is to free up human contributors to focus more on quality control and deliberation. While Wikipedia already uses AI for functions like vandalism detection and translation, this is the first time AI tools will directly assist editors.

The Foundation emphasizes a human-centered, transparent, and open-source approach to AI integration, especially amid growing concerns that AI bots are overwhelming Wikipedia’s infrastructure by massively scraping content, prompting efforts to create a structured, machine-learning-friendly dataset to manage this impact.

BIG TECH

Visa wants to give AI agents your credit card

Visa is partnering with leading AI developers—including OpenAI, Microsoft, Anthropic, Perplexity, and Mistral—to enable AI agents to make purchases on users’ behalf by integrating them directly with Visa’s payment network. These agents, envisioned as more capable successors to current chatbots, could handle tasks like buying groceries or booking travel based on user-defined budgets and preferences. Visa believes this shift could be as transformative as the advent of e-commerce and has spent months working with AI developers to overcome technical and security challenges.

Consumers would maintain control through spending limits and purchase approvals, with the potential for agents to gradually gain autonomy. By allowing AI to access transaction histories (with consent), the system could also personalize recommendations. Though Visa acknowledges AI won’t replace all shopping experiences, it sees a growing role for agents in automating routine or complex purchases, offering convenience while preserving trust and control.

BIG TECH

Amazon launches its most capable AI model

Amazon has launched Nova Premier, its most advanced AI model in the Nova family, capable of processing text, images, and video (but not audio), and now available through its Bedrock platform. With a massive context window of 1 million tokens (about 750,000 words), Nova Premier is designed for complex, multi-step tasks requiring deep contextual understanding, though it underperforms on some coding and STEM benchmarks compared to models like Google’s Gemini 2.5 Pro.

It excels in knowledge retrieval and visual comprehension and is positioned for use in model distillation—training smaller, more efficient models for specific applications. While not a "reasoning" model, it’s competitively priced at $2.50 per million input tokens and $12.50 per million output tokens. Amazon is integrating AI more deeply into its strategy, with CEO Andy Jassy noting rapid AI revenue growth and the development of over 1,000 generative AI applications.