Hermes vs. OpenClaw, Cybersecurity Alarms Ring, More-Interactive Conversations, Can Agents Do Human Work?

2026-05-22 · Source: The Batch | DeepLearning.AI · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Robotics & Autonomous Systems, Cybersecurity & Data Privacy · Depth: Advanced, long

Summary

The latest intelligence brief highlights several key AI advancements and challenges. Nous Research's open-source Hermes Agent, launched in February, has surpassed OpenClaw on OpenRouter's token consumption leaderboard, demonstrating advanced memory architecture and automatic skill creation. Thinking Machines Lab introduced TML-Interaction-Small, a 276-billion-parameter multimodal system that processes audio, video, and text concurrently, achieving 0.40-second response times on FD-bench V1 for real-time, interruptible conversations. Concurrently, Google reported an AI-generated script bypassing two-factor authentication, signaling a rise in industrial-scale cyberattacks and the potential for LLMs to find unknown vulnerabilities. A study from Carnegie Mellon and Stanford Universities revealed that current AI agent benchmarks, based on 10,000 examples from 43 benchmarks, heavily favor software engineering tasks (8,622 examples) over broader economic activities like office administration (3,186 examples) or management (676 examples), indicating a mismatch with the general labor market.

Key takeaway

For AI Engineers developing autonomous systems, you should prioritize integrating self-improving capabilities and robust memory architectures, as seen in Hermes Agent, to enhance agent performance and adaptability. MLOps Engineers deploying conversational AI should explore multimodal systems like TML-Interaction-Small for real-time, natural user experiences, but be mindful of the computational demands. Cybersecurity professionals must urgently address the escalating threat of AI-generated attacks by focusing on proactive defensive research and advanced vulnerability detection, as LLMs can exploit flaws faster than patches are deployed.

Key insights

AI agents are rapidly evolving with self-improvement and real-time interaction, while also posing new cybersecurity risks and revealing benchmark biases.

Principles

• Agent self-improvement enhances capability.
• Concurrent multimodal processing enables natural interaction.
• AI benchmarks should reflect broader economic value.

Method

Hermes Agent uses an agentic loop with automatic skill creation and a Curator system for skill management. TML-Interaction-Small pairs a fast interaction model with an asynchronous background model, interleaving micro-turns for concurrent processing.

In practice

• Hermes Agent supports local or cloud LLM deployment.
• TML-Interaction-Small enables live translation and proactive interjection.

Topics

• AI Agents
• Multimodal AI
• Conversational Systems
• Cybersecurity AI
• AI Benchmarking

Code references

• NousResearch/hermes-agent
• plastic-labs/honcho
• OpenBMB/MiniCPM-o

Best for: CTO, VP of Engineering/Data, Director of AI/ML, AI Scientist, AI Engineer, MLOps Engineer

Related on AIssential

• HERMES AGENT SETUP: the OpenClaw killer is here — Hermes Agent offers continuous self-improvement and skill generation for AI agents, emphasizing local, open-source model deployment.
• Do Multimodal Agents Really Benefit from Tool Use? A Systematic Study of Capability Gains — Tool-augmented multimodal agents may learn tool-calling patterns more reliably than actual tool-contributed capabilities.
• Claude Opus 4.6 and GPT-5.3-Codex 30 Mins Apart — AI agent teams are rapidly advancing, demonstrating autonomous complex task execution and self-improvement capabilities.
• The TechBeat: We brought Hermes Agent to iMessage, even on Linux and Windows (6/25/2026) — The tech landscape is rapidly evolving with AI-driven innovations and infrastructure challenges across diverse domains.
• not much happened today — AI development is rapidly converging on composable agentic systems, multimodal understanding, and efficient local inference.
• Perplexity Max Debuts Multi-AI Agent Tool — Perplexity's new agent tool unifies 19 AI models for complex workflows, targeting professionals with a multimodal strategy.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by The Batch | DeepLearning.AI.