Harrison Chase’s Post

“AI agents” are here right now… Very smart techniques are rapidly evolving to make autonomous “planning” and “reasoning” possible… BUT these only work when having access to the right data and at the right time… The “HOW” to make it possible at an Enterprise Production grade, with the right level of security and governance is THE question… And we, at Synapsewerx have an answer!!! Interesting months ahead…

🧠Planning for agents One of the big limitations for agents working reliably is their ability to plan. We talk about that in a new blog post. Covers: - What is planning? - How to improve planning? - Generic vs custom cognitive architectures and more! https://lnkd.in/gX9cjBGF

Boost Your Agent's Brainpower: Planning Made Easy Agents need to plan! This blog breaks down how to improve their planning skills, covering everything from basic planning to advanced cognitive architectures. Follow for more #AI updates!

Knowing which step to take next is the foundation of an LLM agent. This is what planning is all about. The key in our case was task decompositon, with each function having clear and achievable subgoals. The orchestrator needs a deep understanding of all functions. LangGraph enhances control over this decomposition, as the orchestrator can be made aware of all available nodes and subgraphs and their roles. This allows for transparant planning.

Contract testing focuses on the message itself, not the underlying technology, making it adaptable to various event-driven architectures. Hear more from Lewis Prescott in episode 184 https://lnkd.in/d-4GjJpH

Just a reminder, that my course of the foundations of messaging, Practical Messaging, is running next at NDC London in 2025. - Day 1: Messaging Patterns. We will cover the fundamentals of using messages for interoperability in distributed systems - Day 2: Conversation Patterns. How to compose the building blocks of Day 1 into Event Driven Architectures. Takeaways: how to write messaging code; how to compose event driven architectures See the link in the comments.

How people looks at me when I start talking about integration of the latest technologies in existing architectures:

🛃Generic vs custom cognitive architectures To improve the planning ability of an agent, you may try to improve the cognitive architecture. You can do this in a generic or custom way Agents in production that we see have custom cognitive architectures https://lnkd.in/gwi3WAK3

🛃Generic vs custom cognitive architectures To improve the planning ability of an agent, you may try to improve the cognitive architecture. You can do this in a generic or custom way Agents in production that we see have custom cognitive architectures https://lnkd.in/gX9cjBGF

cloudsa : Forrester reports that over 63% of organizations are struggling to implement #ZeroTrust Architectures, with lack of talent being a key issue. Check out this blog to learn what ZT knowledge your team needs to know & how the CCZT can help → https://bit.ly/49YfHif — CloudSecurityAlliance (@cloudsa) Apr 30, 2024