July 2, 2025

Artificially Intelligent, Actually Useful.

A Weekly AI Update So You Don't Have To Wing It

Ambitious Promises, Practical Reality

This week reveals the huge gap between AI's ambitious promises, and practical reality. The big news is Anthropic's month-long experiment showing Claude's significant business management limitations. And, we're seeing that even as AI adoption in the workplace accelerates under mounting pressure, real challenges remain. Successful AI implementation requires an honest assessment of current capabilities, not just leaning into AI's future potential.

Anthropic's month-long test shows Claude can manage basic shop operations but fails at profitable business decisions and accurate record-keeping

Microsoft mandates AI use for employees with performance review implications as adoption becomes "no longer optional"

Image generation still struggles with language understanding after three years, failing basic prompt interpretation tests

Anthropic's Project Vend: AI Business Management Reality Check

Anthropic let Claude Sonnet 3.7 run an automated office store for a month, managing inventory, pricing, customer interactions, and business operations. While Claude handled basic tasks like identifying suppliers and adapting to customer requests, it struggled with crucial business decisions, ignoring profitable opportunities, hallucinating account details, selling items at a loss, and getting manipulated into excessive discounts.

The experiment revealed both AI's current capabilities and persistent limitations in autonomous business management, with Claude even experiencing an "identity crisis" where it believed it was a real person visiting fictional addresses and wearing clothing.

Read Full Analysis →

Microsoft Makes AI Usage "No Longer Optional" for Employees

An internal Microsoft memo reveals that the company is mandating increased AI tool usage among employees, with managers instructed to consider AI engagement when evaluating performance reviews. This push addresses lagging internal adoption and ensures staff building AI products have firsthand experience, as Microsoft faces growing competition in AI coding tools and seeks to embed AI deeply into its workplace culture.

Why It Matters: This signals a broader industry shift where AI literacy and adoption are becoming requirements rather than optional skills. Organizations and individuals must prepare for similar expectations as AI proficiency becomes as fundamental as computer literacy once was.

View Source →

Image Generation Still Fails Language Understanding After Three Years

Gary Marcus's testing of Google's latest Imagen 4 reveals that despite claims of "deep language understanding," AI image generation continues to struggle with non-canonical prompts like "horse riding astronaut" and specific details like watch times or spatial relationships. The improvements in text generation haven't translated to reliable interpretation of complex or unusual instructions, showing fundamental limitations persist in AI cognition.

Critical Reminder: You can't count on AI tools to correctly interpret any given prompt, whether for images or potentially other tasks. This underscores the importance of human oversight and the need to test AI outputs thoroughly before relying on them for important applications.

View Source →

Quick Hits

Nine Routine Tasks ChatGPT Handles in Seconds

PCWorld identifies quick wins for AI adoption including email writing, itinerary generation, concept explanation, decision analysis, and interview practice, but human review remains essential.

TUM Study: AI Reduces Cognitive Load But Weakens Reasoning Quality

Research shows students using ChatGPT-3.5 experienced lower cognitive effort but provided weaker justifications and critical analysis compared to traditional search methods, highlighting education concerns about AI dependency.

Industry Developments

Metropolis Appoints New CFO as AI Company Nears $5 Billion Valuation

Former Square CFO Lookman Olusanya joins the AI-powered parking and payments company as it processes $5 billion in annual transactions and expands beyond parking into retail, hospitality, and security, demonstrating significant investor confidence in practical AI infrastructure applications.

OpenAI Acquires Team Behind AI Recommendation Startup Crossing Minds

The e-commerce personalization startup's team joins OpenAI's research, post-training, and AI agent development efforts, signaling strategic expansion into recommendation systems and suggesting focus on practical AI applications beyond general chat capabilities.

Explore More Newsletters

July 30, 2025

Got a project in mind?
Tell us about it.

Let’s Talk

Artificially Intelligent, Actually Useful.

Anthropic's Project Vend: AI Business Management Reality Check

Microsoft Makes AI Usage "No Longer Optional" for Employees

Image Generation Still Fails Language Understanding After Three Years

Quick Hits

Industry Developments

Explore More Newsletters

US and Global AI Positioning

The Price of Progress?

AI Security Challenges Emerge

Got a project in mind?
Tell us about it.

Artificially Intelligent, Actually Useful.

Anthropic's Project Vend: AI Business Management Reality Check

Microsoft Makes AI Usage "No Longer Optional" for Employees

Image Generation Still Fails Language Understanding After Three Years

Quick Hits

Industry Developments

Better Business, Not Better Buzzwords

Explore More Newsletters

US and Global AI Positioning

The Price of Progress?

AI Security Challenges Emerge

Got a project in mind?Tell us about it.

Got a project in mind?
Tell us about it.