Cobus Greyling

1 week ago | cobusgreyling.medium.com | Cobus Greyling

Phi-4 Language Model open-sourced by Microsoft — Synthetic data, AI Agents, Synthetic Generation Prompts and Hallucination Refusal. In training Phi-4, a synthetic data pipeline was used to transform factual snippets into engaging, multi-turn conversations that flow naturally and promote deep reasoning. Initial ResponseThe model processes a factual snippet and generates a clear, foundational response that conveys the core information.

1 week ago | cobusgreyling.medium.com | Cobus Greyling

Lessons from Web & Voice Device Discovery to Enhance AI Agent Accessibility. The web had a problem with discoverability…websites where out there, but we struggled to find them. So, the notion of a search engine was conceived.

1 week ago | cobusgreyling.medium.com | Cobus Greyling

Agentic workflows are changing how knowledge workers are interacting with technology. Workflows use autonomous AI Agents to perform complex tasks with minimal human intervention, but oversight. While AI Agents introduce a new level of autonomy, they still require strong human supervision to ensure accuracy and reliability. Agentic workflows refer to processes where AI Agents — powered by advanced large language models (LLMs) — execute multi-step tasks autonomously.

1 week ago | cobusgreyling.medium.com | Cobus Greyling

Dynamic Self-Discovery with MCP Let Your LLM Find Its Own ToolsLarge Language model (LLM) based applications are evolving from passive assistants to actively exploring their digital environments to fulfil user intent. At the beginning, LLMs only had access to their trained knowledge base, this changed when In-Context Learning (ICL) was discovered. Which spawned a whole RAG revolution.

2 weeks ago | cobusgreyling.medium.com | Cobus Greyling

How MCP Servers Supercharge vertical AI Agent IntegrationPersonally, I liked the whole idea of function calling, allowing large language models (LLMs) to interact with external tools, APIs, and data sources to perform tasks beyond their training data. Function calling bridged a gap, enabling AI to execute predefined actions.