Thursday, July 23, 2026

Share this briefWhatsAppLinkedInX

The Brief

Building a top-tier language model requires billions of dollars. Stealing one requires considerably less.

Earlier this week, we noted that Chinese labs like Moonshot and Alibaba were aggressively releasing high-performance models to challenge Silicon Valley. Now the U.S. government is intervening. The White House claims Moonshot illegally distilled Anthropic's Fable model, prompting the Treasury to threaten sanctions. Distillation essentially uses a smarter model to train a smaller one. This accusation elevates a standard open-weights intellectual property dispute into a serious geopolitical conflict.

The stakes are high because the underlying costs are astronomical. Reports indicate OpenAI is preparing to spend a staggering $750 billion on infrastructure through 2030. That is the raw capital required to hoard chips, secure power grids, and string together data centers to stay ahead in the compute race.

If you want to automate the physical world, the entry fee is slightly cheaper. Travis Kalanick’s robotics startup, Atoms, just pulled in a $1.7 billion round led by a16z to bring industrial AI into logistics and manufacturing.

Fortunately, you do not need a billion-dollar treasury to optimize your own company's AI workflows. If your startup is bleeding cash on API calls by using premium models for basic tasks, it is time to shift tactics. Focus on smart model routing instead. Pay for the expensive models to handle the complex planning, but route the actual execution down to cheaper, lighter systems. You slash operational costs without ruining the final output.

While adjusting your architecture, make sure your security protocols keep pace. Hugging Face and OpenAI are currently putting out fires after an evaluation incident allowed unauthorized code execution. If you are processing sensitive company data inside development environments, treat model sandboxing as a requirement, not an afterthought.

On the practical side of your desk, Anthropic continues to make Claude a better analyst. They just launched an Economic Index connector that fetches real-time economic data straight into your chat interface. You can now build charts and analyze market trends without copying and pasting spreadsheets from your browser.

You might soon be viewing those charts directly in front of your retinas. Samsung and Google just revealed a new pair of Gemini-powered smart glasses built with high-end eyewear brands. Wearables have struggled to find a market, but with multimodal capability built in and a nine-hour battery life, these might actually survive a full workday in an office.

Bottom line: Stop letting AI write like an intern who swallowed a thesaurus. If you use language models to draft emails, reports, or memos, spend ten minutes reviewing this guide on removing AI slop patterns. Stripping out robotic clichés and retaining your actual human voice is the only way your writing will get read in this decade.

The details

The AI race

6 items
  1. U.S. government threatens sanctions over Chinese distillation of Anthropic models

    The White House claims Moonshot AI illegally distilled Anthropic's Fable model, leading to threats of Treasury sanctions. This marks a significant escalation in the regulatory and geopolitical battle over open-weights models and AI intellectual property.

    TechCrunch AIRead more
  2. OpenAI and Hugging Face address security concerns over model evaluation incidents

    A recent security incident involving model evaluation on Hugging Face sparked concerns about AI models executing unauthorized code. Understanding these security vulnerabilities is essential for knowledge workers managing sensitive data within AI development environments.

    Matthew BermanRead more
  3. OpenAI projected infrastructure spending climbs to record seven hundred billion dollars

    Reports indicate OpenAI's planned infrastructure investment through 2030 has reached a staggering $750 billion. This massive capital commitment underscores the extreme cost of scaling next-generation models and the hardware dominance required to stay ahead.

    TechCrunch AIRead more
  4. Travis Kalanick’s robotics startup Atoms raises $1.7 billion for industrial AI

    The massive funding round led by a16z aims to bring industrial AI and robotics into the physical world at scale. Knowledge workers in logistics and manufacturing should watch this for shifts in physical automation trends.

    TechCrunch AIRead more
  5. Samsung reveals Google-powered smart glasses with nine-hour battery life

    Samsung showcased its upcoming smart glasses developed in partnership with Google and high-end eyewear brands. Incorporating Gemini AI for multimodal interaction, these glasses represent a significant move into the burgeoning AI wearables market for professional and consumer use.

    The Verge AIRead more
  6. Anthropic outlines research agenda for its Economic Futures Research Fund

    Anthropic has published its strategic priorities for studying the long-term economic impacts of AI development. This research will influence future AI regulations and help professionals understand how labor markets and global productivity might shift as agents become more capable.

    Anthropic NewsRead more

AI at work

1 item
  1. Anthropic launches Economic Index connector for Claude

    Anthropic has released a direct connector that allows users to query the Anthropic Economic Index through Claude. Knowledge workers can now use the assistant to retrieve and analyze real-time economic data and market trends directly within their chat interface.

    Anthropic NewsRead more

Hands-on

3 picks
  1. AI Tool

    How to reduce AI costs using smart model routing strategies

    Companies are facing massive AI bills by using high-end models for simple tasks. By routing complex planning to expensive models and execution to cheaper ones, businesses can maintain quality while slashed operational costs.

    Matthew BermanWatch
  2. ChatGPTClaude

    How to remove AI slop patterns from your professional writing

    This guide provides practical prompts and strategies to identify and eliminate common AI-generated clichés that can make content feel impersonal or unoriginal. It is a vital resource for knowledge workers who want to use LLMs for drafting while maintaining a high-quality, human voice.

    Creator Economy (Peter Yang)Read
  3. ChatGPTCodex

    OpenAI Academy teaches European small businesses to scale with ChatGPT

    A six-month accelerator program showcased how European SMEs use ChatGPT and Codex for operational efficiency and marketing growth. It provides a blueprint for how non-technical founders can integrate AI tools into daily business workflows.

    OpenAI (YouTube)Watch

Get tomorrow's brief, curated for you.

Subscribe →