Hi all,

It's already April! This year is going by quickly and there are a lot of developments in the agent space. To that end, I'm working on the first edition of an agent tech radar - tracking the tools and practices used in practice.

A bonus email is going out tomorrow with all content that didn't fit here.

Cheers,
Ivan

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

Events

• Apr 3 - The Agentic AI Mixer – Finance & Tech Edition
• Apr 3 - Building Agents: NYC Hackathon
• Apr 4 - Personalized Agents Hackathon
• Apr 8 - Building AI Agents That Actually Remember
• Apr 8 - OpenClaw for Life and Business: AI Agents & GrowthOps
• Apr 8 - Tech Talk: Building Agentic Media Buying Platforms
• Apr 9 - Building Agents: Designing, Shipping, and Controlling the Cost of AI
• Apr 9 - OpenClaw Meetup NYC
• Apr 10 - Enterprise Agents Hackathon
• Apr 11 - Intro to IronClaw: Agentic Workshop
• Apr 11 - Building Your Own AI Agent: A Guided Walkthrough
• Apr 16 - Agentic Knowledge Meetup
• Apr 16 - NYC Product Leaders Breakfast: Let the AI Agents Work (but on what?)
• Apr 16 - LangChain Presents: AI Agents Workshop with Harrison Chase
• Apr 16 - E-commerce in the Age of Agents
• Apr 17 - AWS & zeb - Build Smarter: Deploying Production-Ready AI Agents on Amazon Bedrock AgentCore (NYC)
• Apr 17 - No-Code AI Agent Workshop with MuleRun Ambassadors
• May 4 - AI Agent Conference

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

News

• AI agents could easily send college grad unemployment over 30%, ServiceNow CEO says
• Gartner Predicts at Least 80% of Governments Will Deploy AI Agents
• Google finds that AI agents learn to cooperate when trained against unpredictable opponents
• The rise of AI agents tests Beijing’s playbook
• Mark Zuckerberg Is Building an AI Agent to Help Him Be CEO

The headlines range from ominous to frivolous, but they are definitely more consequential when compared to those from last year. Agents are no longer theoretical, they're real and important.

Launches

• Claude computer use
• Stripe Projects: provision services from the command line
• NVIDIA announces NemoClaw
• Introducing Wiz Agents & Workflows: Security at the Speed of AI
• Introducing My Computer: When Manus Meets Your Desktop
• Agentcard: virtual cards for agents
• Introducing LangSmith Fleet
• Introducing Perplexity Computer

Big companies keep shipping agent products and features relentlessly. Agent computers are the current trend, but it's dubious the market needs dozens of offerings. Stripe's Projects product is also a really interesting development for businesses.

Deals

• Harvey ($200M) - legal AI agents
• Isara ($94M) - agent swarm coordination
• RunSybil ($40M) - offensive security agents
• Notch ($30M) - customer service agents
• Interloom ($16M) - agent workflow understanding
• CometChat ($6M) - conversational AI agents
• Respan ($5M) - proactive agent observability

Tons of funding going out these days - lots going to companies that are just coming out of stealth. Many of these startups I'm hearing for the first time even as they're closing a Series A round.

Articles

• Agents Over Bubbles by Ben Thompson
• Legal is Next by Harvey
• Agents for Security: The Tipping Point for Offensive AI by Menlo Ventures
• Agent Computers: The PC Era, Amplified by AMD
• What we wish we knew about building AI agents by PostHog
• Securing AI agents: the defining cybersecurity challenge of 2026 by Bessemer Venture Partners
• After all the hype, was 2025 really the year of AI agents? by Stack Overflow

Industry analysts and investors are pushing out thinkpieces to recalibrate expectations for this year after a very optimistic 2025. The questions now are: what will the next generation of agents look like, and where will they go?

Projects

• autocontext - self-improving harness
• AgentEvals - evaluators for agents
• JAI - jail for agents

Infrastructure is becoming the area of focus in open source agent projects. I think the stack will evolve actively for a long time until settling down.

Learning

• Agent Skills with Anthropic
• Evaluating AI Agents in Practice: Benchmarks, Frameworks, and Lessons Learned
• Production-Ready LLM Agents: A Comprehensive Framework for Offline Evaluation
• How we build evals for Deep Agents
• Background Agents

Evals are on everyone's mind lately and so a good amount of learning material has come out to help others understand what to do and how to do it.

Research

• Hyperagents
• τ-bench
• AI Agent Traps
• Automating Skill Acquisition through Large-Scale Mining of Open-Source Agentic Repositories
• Agentic AI and the next intelligence explosion
• Natural-Language Agent Harnesses
• The Evolution of Tool Use in LLM Agents
• YC-Bench: Benchmarking AI Agents for Long-Term Planning and Consistent Execution
• Drop the Hierarchy and Roles: How Self-Organizing LLM Agents Outperform Designed Structures
• Securing LLM agents: From prompt sanitization to autonomous red teaming and beyond

There is a ton of research in this edition with a lot of really interesting papers. Subjectively it feels like research has really picked up lately - I'd guess it's because many practical problems have been discovered in the last year.

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