Vladimir Stajilov
Picking a Memory Engine for an Agent Loop
An agent without memory is a very expensive goldfish. We compared four memory engines for our Whale harness — Semantica, the Neo4j agent memory layer, OpenWolf and cognee — and the four criteria that decided it.
Procedural Graphs: Memory That Rewrites Itself
Most memory stores facts — what is what. A procedural graph stores what to do, and rewrites itself when a step fails. How we implemented it in Whale, and the three decisions the paper does not make for you.
An Agent That Plans Needs an Adversary and a Human Gate
A bad single action is cheap. A good plan pointed in the wrong direction is expensive. We put a critic and one human approval in front of our agent’s plan — and the first test run showed exactly why.
Creating an Agent Harness: From the Loop to a Multi-Agent Architecture
Most of what gets sold as an “AI agent” is a single model call in a trench coat. We are building a real one in public. Whale: the loop that makes agents click, and the architecture that goes beyond ReAct.
Mastering Digital Sovereignty: Strategies for IT Control and Transparency
In a world where technology evolves at an unprecedented pace, businesses face the challenge of maintaining control, transparency, and efficiency in their digital infrastructure. At nowtec solutions, we’ve taken significant strides toward achieving digital sovereignty – a concept that emphasizes…
Strapi-tacular Tweaks: A Deep Dive into Custom Plugins and Advanced Features
Content management systems (CMS) have become an integral part of modern web development, and Strapi stands out as a powerful, flexible solution for businesses of all sizes. At nowtec solutions, we’ve embraced Strapi as our go-to CMS, and we’re constantly…
LLMs: Year in Review, Mastering the Art of Large Language Models
As we bid farewell to 2024, it’s clear that this year has been a game-changer in the world of Artificial Intelligence, particularly in the realm of Large Language Models (LLMs). At nowtec solutions, we’ve been at the forefront of this…
Exploring Private LLMs for Data Privacy and Control
Private Large Language Models (LLMs) are revolutionizing the way businesses handle data, offering a secure and transparent alternative to proprietary AI solutions. By running open-source LLMs on private infrastructure, companies can maintain full control over their data while leveraging the…
WordPress Security: Tackling the WP Time Capsule Vulnerability
The WordPress ecosystem recently faced a significant security challenge with the discovery of a critical vulnerability in the popular WP Time Capsule plugin. This article delves into the issue, its impact, and provides a solution for affected websites. The Vulnerability…
Private AI: Exploring Open-Source LLMs for Secure Enterprise Solutions
Private AI is transforming the way businesses handle sensitive data, and at nowtec solutions, we’re at the forefront of this revolution. Our recent exploration of open-source large language models (LLMs) has led us to an exciting discovery: Ollama models are…