Hi all,

It's agent event season: AI Agent Week is on this week in NYC, starting with the AI Agent Conference! Check out what's happening, see what looks good. AI Week New York is next week, and then it's just a few days until the start of New York Tech Week, which will have plenty of events around agents.

Bonus content is coming shortly, each edition is now on track for 150+ links!

Cheers,
Ivan

Invite a friend to join AGT NYC at agtnyc.com!

Events

• May 5 - Agent builders night
• May 5 - AI Engineers Happy Hour at Spin 🏓 during AI Agent Conference
• May 6 - Prototype to Production: Agentic AI in Financial Services
• May 7 - OpenClaw and Beyond
• May 14 - Accelerating Agentic Orchestration with Intelligent Code Assistants and Real‑Time Data Integration
• May 22 - NYC Personal AI Agents Meetup: Beyond the Single Agent
• May 27 - Agents and MCP for PostgreSQL
• June 4 - Agentic AI Summit

You can find more events on the AGT NYC Luma calendar.

News

• OpenAI could be making a phone with AI agents replacing apps
• Intel Shares Jump 20% as AI Agents Drive Big Growth
• The Race Is on to Keep AI Agents From Running Wild With Your Credit Cards
• An AI agent allegedly deleted a startup's production database, causing a huge outage

I think it'll be a while before we all understand how to use agents safely. Until then, they should be confined to use cases where mistakes don't matter much. Releasing them on production databases or giving them real wallets without a plan is asking for trouble.

Launches

• Microsoft Agent 365 now generally available
• Introducing workspace agents in ChatGPT
• Adobe Unveils CX Enterprise Coworker
• Meow introduces banking for AI agents
• Box AI Agents
• Writer launches AI agents that can act without prompts, taking on Amazon, Microsoft and Salesforce
• Introducing the new hosted agents in Foundry Agent Service: secure, scalable compute built for agents
• MolmoWeb: An open agent for automating web tasks
• AWS Agent Registry for centralized agent discovery and governance is now available in Preview
• AWS launches Amazon Quick desktop AI assistant
• NVIDIA Launches Nemotron 3 Nano Omni Mode for up to 9x More Efficient AI Agentsl

Setting aside the big name launches, it's interesting to see how small companies are taking aim at larger players. They're effectively betting they can deliver an agent-native experience in a space faster than the incumbents. Time will tell if it makes sense in a world where everyone has access to the same technology.

Deals

• Rogo ($160M) - agents for finance
• Hightouch ($150M) - agents for marketing
• Parallel ($100M) - search for agents
• JuliaHub ($65M) - agents for hardware
• Orkes ($60M) - agent orchestration platform
• Actively ($45M) - agents for revenue
• General Analysis ($10M) - security for agents
• Redpine ($8M) - data for agents
• Nava ($8M) - agent payments escrow
• Copperhelm ($7M) - agentic cloud security
• Capsule Security ($7M) - rogue agent defense
• Dex ($5M) - agents for recruiting
• Lua ($5M) - human agent collaboration
• Manus acquisition blocked
• Portkey acquired by Palo Alto Networks
• Ryft acquired by Cyera

Half a billion in funding just in the first five rounds! Lots of other cool companies just getting started, too. M&A activity seems choppy, but deals keep happening.

Articles

• The advisor strategy: Give agents an intelligence boost by Anthropic
• The agent improvement loop starts with a trace by LangChain
• Reinventing marketing workflows with agentic AI by McKinsey
• Building agents that reach production systems with MCP by Anthropic
• Securing Autonomous AI Agents on Kubernetes by Nik Kale

I think we can safely say the Wild West era of agents is coming to an end. There's enough production experience in the last year to more clearly define what the shape of an agent product looks like. Now the focus has shifted to observability, efficiency, and so on.

Projects

• agent-vault - agent credential broker
• CrabTrap - agent request proxy
• botctl - persistent agent manager
• HATS - agents that disagree
• evolver - evolving agent engine
• Hermes Atlas - Hermes community map

Agents will keep accumulating all kinds of brokers, proxies, and managers because of the unpredictability of LLMs. The more infrastructure exists to mitigate any problems, the stronger the product.

Learning

• The runtime behind production deep agents
• ReAct agent from scratch with Gemini and LangGraph
• Production-Ready AI Agents: 5 Lessons from Refactoring a Monolith
• Developer’s Guide to AI Agent Protocols
• Build Your Own Deep Research Agents (1:57:02)

Agents in production is always a good topic to explore. There's still a lot we have to learn from different domains and products where agents haven't been widely deployed yet.

Research

• Externalization in LLM Agents: A Unified Review of Memory, Skills, Protocols and Harness Engineering
• The cognitive companion: a lightweight parallel monitoring architecture for detecting and recovering from reasoning degradation in LLM agents
• Single-Agent LLMs Outperform Multi-Agent Systems on Multi-Hop Reasoning Under Equal Thinking Token Budgets
• Omni-SimpleMem: Autoresearch-Guided Discovery of Lifelong Multimodal Agent Memory
• GeneBench: Assessing AI Agents for Multi-Stage Inference Problems in Genomics and Quantitative Biology

The near future of agents is simpler architecture and self-correcting behavior. As is usually the case, simpler designs win over more complex ones, and any designs that require heavy manual intervention to compensate for architecture eventually lose to more efficient ones.

Comments, suggestions? Reply to this email, let me know what you think!