Hi all,

Hope everybody had a great summer! This is the second AGT NYC newsletter, with some more resources I hope you find useful.

Some group insights from the event poll so far:

• Email is by far the most preferred way to communicate
• People would prefer to meet once or twice a month
• There is a strong preference to meet in person
• Midtown is #1 for meeting location, #2 is Flatiron/USQ
• People prefer to meet Monday evening (alt: Tue eve and Fri lunch)

Still putting together plans for the first event. Any referrals to a venue or event sponsor would be greatly appreciated.

If you have any agent news, projects, posts, videos, or papers, let me know!

News

• Epic, the healthcare company, deploys first agents
• Workday acquires Flowise, the AI agent builder
• NIH develops AI agent to improve gene set analysis
• Databricks acquires Tecton to personalize AI agents
• Stanford Medicine created a virtual lab with AI agents

Agent-powered science is an area to watch - it looks like the most valuable application of agents next to business processes. As far as the recent agent-related deals, there will probably be more incumbents doing this to gain a lead in AI. The cool thing about the agents space is that nobody really owns it and there is a lot of value up for grabs.

Recent fundraising: E2B ($21M), Composio ($25M), FRL ($30M)

Posts

• Agents Built from Alloys from XBOW
• The state of AI agents by Lance Martin
• Why I'm betting against AI agents by Utkarsh Kanwat
• Simulations: the secret behind every great agent from Sierra
• Don’t let hype about AI agents get ahead of reality in Technology Review

Since 2025 is supposed to be the "year of agents," I think it's interesting that it's seen such a measured and pragmatic response from industry experts and practitioners. People recognize there's something valuable here and don't want the hype cycle to sink promising and useful technology.

Videos

Making AI Agents Work For You (and Your Team)

Hannah Foxwell discusses the design and implementation of AI agent teams, arguing that agents should be used to automate "toil" and reshape organizational structures.

LangChain: 3 ingredients for building reliable enterprise agents

Harrison Chase covers ways to make agents more successful in enterprise, including adding determinism, using observability and evals, and making decisions reversible and supervisable.

Why Devin does not use multi agents

Useful conversation between two practitioners about the limitations of and alternative architectures to multi-agent systems.

LangChain: Context Engineering for Agents

Find out how to optimize performance and prevent failures with context engineering techniques like context compression and isolation, scratch pads, and more.

LlamaIndex: Effective agent design patterns in production

Laurie Voss explores patterns for real-world production agents, including chaining, routing, and self-reflection.

Projects

• AGENTS.md - a simple, open format for guiding coding agents
• Parlant - structure for agents in natural language
• any-agent - single interface for different agent frameworks
• Strands Agents - model-driven approach to building AI agents
• Airweave - agent search interface for any app

I'm getting the feeling we're entering a second wave of agent tooling, where the framework wars are less important than getting the best results. This second wave may emphasize interoperability, standards, and less reliance on specific code and tools. For agents, personal preference for particular tools might be trumped by the setup that wins.

Papers

• Chain-of-Agents
• A Comprehensive Survey of Self-Evolving AI Agents
• AI Agents and Agentic AI...for Future Manufacturing
• Future of Work with AI Agents
• The Term 'Agent' Has Been Diluted Beyond Utility and Requires Redefinition

Some exciting stuff in every paper - as for the last one, I think it's long overdue. It remains to be seen just how much (true) agents will affect the real world, but the research around agent evolution and deploying agents in industry is exciting. The next frontier, agents that develop themselves and other agents, is slowly coming into view.

​If you know anybody else that would be interested in AGT NYC, please refer them to agtnyc.com to sign up.

Cheers,
Ivan