GenAI/AI News Jun 30 25: AI is Gobbling Up R&D Budgets, The Modes of Agentic Attack Are Articulated, and Much More...

ICONIQ's survey helps show where money is being spent by innovators, and a wonderful compendium of AI communication methods, risks and mitigations from a team of collaborators from China, Saudi etc.

I love facts. This new survey from iconic helps layout how AI is eating R&D budgets. The vast majority of new companies in the AI space are building around agentic systems and 47% of new AI firms have achieved product market fit and more. In the second essential article there's a wonderful compendium of AI agent communications put together in a framework that articulates risks and mitigations. Looks like an exciting start to our week.

Good Day, GAI Insights Community! If you missed us today, watch today's show on YouTube. Join the fun! Watch and comment on our Live feed on Linkedin or YouTube. We post articles for the daily briefing here so you can follow along and comment. Today there were 2 Essential, 2 Important, and 2 Optional articles.

2025 State of AI Report: The Builder’s Playbook

Rating: Essential
Rationale: ICONIQ Growth's report synthesizes insights from 300 software executives into a dense, visual playbook on how to build, scale, and structure AI-driven products and companies. Our analysts unanimously flagged this as a vital, strategy-oriented guide for AI leaders, investors, and enterprise builders due to its breadth, clarity, and relevance across key implementation fronts.

A Survey of LLM-Driven AI Agent Communication

Rating: Essential
Rationale: This comprehensive academic survey outlines communication frameworks, security vulnerabilities, and mitigation strategies for large language model (LLM) based AI agents, spanning user-agent, agent-agent, and agent-environment interactions. Our analysts agreed it’s a critical reference for security leaders and AI architects navigating the evolving landscape of agentic AI systems, especially given its relevance to emerging protocols like MCP and A2A.

Project Vend: Can Claude run a small shop? (And why does that matter?)

Rating: Important
Rationale: Anthropic’s Project VEND simulates an autonomous AI system managing a physical store to explore the boundaries of agentic AI in real-world settings. While the experiment highlighted the current limitations in AI contextual understanding, our analysts agreed it represents a meaningful milestone in the development of AI agents and underscores key lessons for future applications.

Boost Embedding Model Accuracy for Custom Information Retrieval

Rating: Important
Rationale: NVIDIA showcases how preparing quality data for small embedding models can significantly enhance custom information retrieval performance, particularly in low-context environments like customer service. Our analysts viewed this as a practical and timely demonstration of how AI accuracy gains can be achieved with curated data, making it a valuable read for technical AI implementers.

Evaluating Long-Context Question & Answer Systems

Rating: Optional
Rationale: Eugene Yan’s long-form exploration of evaluation metrics for long-context Q&A systems offers a thorough conceptual foundation but lacks applied examples. Our analysts noted the theoretical nature of the article limits its immediate utility for AI leaders, though it touches on important issues in AI-driven content retrieval.

Unify engineers growth by using the right model for every task

Rating: Optional
Rationale: OpenAI and Unify describe a high-level overview of using multiple models for distinct tasks like research, observation, and content generation. While the article introduces a useful concept of model specialization, our analysts found it too superficial to warrant a higher rating, especially given the availability of more detailed educational resources.

Like this email? Refer a friend! They can sign up here.

GAI World 2025 Is Coming – Are You Ready? The future of enterprise AI has a date: September 29–30 at the Hynes Convention Center in Boston. GAI World 2025 is where AI leaders, builders, and bold decision-makers collide to share what’s working right now in GenAI. With 800+ attendees, 120 powerhouse speakers, and the most curated AI networking experience available, this is the event where strategies get sharpened, products get launched, and C-suites leave with answers. Want in? Sponsor, exhibit, or attend: www.gaiworld.com

GAI Insights is an industry analyst firm helping AI leaders and their teams achieve business results with GenAl.

Enjoy, John Sviokla, Paul Baier and our analysts, Luda Kopeikina, Adam Rappaport and our Guest Analyst today, Ram Katamaraja. Thank you for joining, Ram!