Hi all,

Stay warm this weekend, maybe build an agent or two. I'll be working on the updated AGT NYC site and test-driving a few AI coding tools while I'm at it. There's a ton of material in this newsletter, hope you find something useful!

Reminder: for the next AGT event, we're looking for presenters and panelists to discuss and demo successful AI projects and what it takes to make them work. If you'd like to get involved, please fill out this form.

Events

• Jan 27 - Agentic AI in the Trenches: Scaling Intelligence for the Real World
• Jan 28 - NYC Voice AI Meetup: The State of Voice Agents
• Feb 11 - Agents & APIs NYC Developer Meetup
• Feb 12 - AI Agents: From PoC to Production - NYC
• Mar 9 - AgentCon New York

News

• Slackbot is an AI agent now
• Claude’s new AI agent pushes down software stocks
• AI Agents Drive First Large-Scale Autonomous Cyberattack
• eBay bans illicit automated shopping amid rapid rise of AI agents
• OpenAI and ServiceNow Strike Deal to Put AI Agents in Business Software
• Google launches Universal Commerce Protocol

The agent moment has definitely arrived - legacy software companies are reinventing themselves around agents or fighting the trend as best as possible. The rest of the economy is likely going to feel a massive shift soon.

Fundraising

• WitnessAI ($58M) - security for agents
• Sequence ($20M) - agents for revenue operations
• Nexxa.ai ($9M) - heavy industry agents
• Tivara ($3.6M) - agents for healthcare admin

It's a slower period than usual for agent startup fundraising, but big deals are still getting done. I expect we'll see fewer startups emphasize agents as time goes on, since they'll become table stakes.

Articles

• Agent-native Architectures by Dan Shipper
• 2026: This is AGI by Sequoia
• A new era of agents, a new era of posture by Microsoft
• Streamlining security investigations with agents by Slack
• Agentic AI advances by McKinsey

If last year everyone was waiting for the other shoe to drop, they don't have to wait any longer. We are maybe 6-12 months away from everyone catching up to the frontier today.

Projects

• Cowork - Claude Code for the rest of your work
• OpenWork - open-source alternative to Claude Cowork
• Skills.sh - agent skills directory
• agent-browser - browser automation CLI for AI agents
• Gambit - agent harness framework
• Mastra 1.0 - agent framework
• agent-sandbox - isolated stateful workloads
• webctl - browser automation via CLI
• HyperAgent - Playwright with AI
• agentgateway - proxy for agents and MCP servers
• GitHub Copilot SDK - agent runtime

The themes for recent launches include: agents for non-technical work, browser automation designed for agents, and the last generation of agent frameworks before the next one.

Learning

• Recursive Language Models: the paradigm of 2026
• How to build agents with filesystems and bash
• Demystifying evals for AI agents
• Agent design patterns
• The complete guide to building agents with the Claude Agent SDK
• Making sense of memory in AI agents
• Building AI agents with just bash and a filesystem in TypeScript
• Choosing the right multi-agent architecture
• Experiences from building enterprise agents with DSPy and GEPA
• Production-grade agentic AI system
• Building Durable, Production Ready Agents (1:18:29)
• From AI agent prototype to product: Lessons from building AWS DevOps Agent
• The agentic AI handbook: production-ready patterns
• Inside Browser Automation: Andrew Baker on Agents, Playwright, and Claude Draws (23:45)

Agents are now in production and a lot of the experience with them so far tends towards simplifying their support systems. Overengineered tools are giving way to older generic approaches that let existing mature systems handle a big part of the workload.

Research

• Agentic Reasoning for Large Language Models
• Introducing APEX-Agents
• The 2026 State of AI Agents Report
• An Empirical Study of Agent Developer Practices in AI Agent Frameworks
• WebGym: Scaling Training Environments for Visual Web Agents with Realistic Tasks
• NeurIPS 2025: 45 Computer-Use Agent Papers You Should Know About
• When Single-Agent with Skills Replace Multi-Agent Systems and When They Fail
• Agent-as-a-Judge
• AI Agent Systems: Architectures, Applications, and Evaluation
• Leveraging LLM-based agents for social science research: insights from citation network simulations

The benchmarks, surveys, and round-ups keep on coming. If last year everyone was still experimenting and prototyping, this year everyone will start putting lessons learned into practice.

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

Cheers,
Ivan