-
By Tom Fenton
Testing small LLMs in a VMware Workstation VM on an Intel-based laptop reveals performance speeds orders of magnitude faster than on a Raspberry Pi 5, demonstrating that local AI limitations are hardware-driven -- not inherent to running AI locally.
-
At TechMentor & Cybersecurity Live! @ Microsoft HQ, Pavan Reddy's "When AI Tools Go Rogue: Securing Agents, MCP, and Dev Assistants" will examine the risks emerging as AI copilots move from chat windows into real systems, and the concrete controls teams need to keep those tools from becoming attack paths.
-
As AI moves from experimentation to production, IT teams are increasingly responsible for making sure it is secure, governable, reliable and cost-effective. This session explores what that looks like in practice, with Eric D. Boyd outlining the architectural, operational and security considerations organizations need to address to run AI at enterprise scale.
-
During Google Cloud Next, Rubrik rolled out one announcement aimed at AI agent governance and another focused on cyber resilience for Google Cloud SQL.
-
AI integration is most effective when you constrain model output through structured prompts and enforce application-side validation so your business logic, compliance requirements, and user experience remain under your control.
-
By Tom Fenton
Benchmarking four compact LLMs on a Raspberry Pi 500+ shows that smaller models such as TinyLlama are far more practical for local edge workloads, while reasoning-focused models trade latency for deeper output quality.
-
Nutanix used its .NEXT 2026 event in Chicago to roll out a broader Nutanix Cloud Platform update and a related expansion of its agentic AI offerings aimed at neocloud providers, adding new infrastructure, multitenancy and management capabilities tied to AI workloads.<
-
By Tom Fenton
Tom Fenton reports running Ollama on a Windows 11 laptop with an older eGPU (NVIDIA Quadro P2200) connected via Thunderbolt dramatically outperforms both CPU-only native Windows and VM-based configurations for local LLM inference -- proving that even modest GPU acceleration delivers significant, real-world performance gains.
-
Stephen L. Rose's TechMentor session offers IT pros a practical, beginner-friendly path into Copilot automation, including agents, prompts, connectors, actions and the governance controls needed to manage them securely.
-
By Tom Fenton
Running an LLM on a Raspberry Pi with Ollama is easy to set up in minutes, but practical performance depends heavily on choosing a smaller model tuned for low-power hardware.
-
By Tom Fenton
This first article in a series explains the core AI concepts behind running LLM and RAG workloads on a Raspberry Pi, including why local AI is useful and what tradeoffs to expect.
-
A hands-on PoC with Microsoft Copilot Studio found that creating a basic agent was easy, but getting useful results from a SharePoint-hosted Excel tracker required extra connection setup and worked better only after separate worksheet tabs were split into distinct files.
-
New agentic AI offering targets release-note review, impact analysis and test planning for enterprise SaaS updates.
-
AWS and Google Cloud used GTC 2026 to detail new NVIDIA-based cloud offerings spanning GPU scale-out, inference, orchestration, and flexible consumption models, while related NVIDIA announcements added context for the company's broader AI cloud strategy.
-
Rubrik announced Rubrik Data Protection for Google Workspace, positioning it as a unified cyber-resilience offering for Gmail and Google Drive with recovery, policy and continuity features.
-
By Tom Fenton
Enterprises face five hard truths when scaling AI from successful pilots to production -- governance gaps, AI agent sprawl, security as an afterthought, agent unpredictability, and the absence of shared architecture -- all lessons that mirror the chaotic early days of API development and demand the same solutions that eventually tamed it.
-
By Tom Fenton
In Tom Fenton's conversation with Solo.io Global Field CTO Christian Posta, the discussion focused on operationalizing AI agents and skills as production-grade, governed resources--including how kagent and agentgateway fit into scaling agent deployments securely across enterprise environments.
-
Veeam unveiled Agent Commander, a unified platform designed to detect AI risk, protect AI systems and precisely undo AI-driven actions at enterprise scale.
-
Druva introduced Deep Analysis Agents, Agentic Memory and multimodal capabilities to move DruAI from conversational assistance to delegated investigative workflows.
-
Rubrik has made Rubrik Agent Cloud generally available, adding expanded governance controls that enforce predefined and custom policies on both AI agent prompts and responses.
-
This walkthrough from Brien Posey shows how to build a custom Ollama model from a base model and tune it with a short Modelfile -- without training from scratch.
-
Greg Schulz explains the critical shift toward API-defined infrastructure and how organizations must leverage federated management and gateways to scale for the unique, high-velocity demands of autonomous agentic systems.
-
Understanding GPU memory requirements is essential for AI workloads, as VRAM capacity--not processing power--determines which models you can run, with total memory needs typically exceeding model size by 20-50% due to weights, activations, KV cache, and system overhead.
-
By Tom Fenton
Tom Fenton details how he used GitHub Pages, PowerShell, and AI-assisted vibe coding to build and maintain a free personal archive of his published articles.
-
If you're a PC Luddite stubbornly clinging to your non-upgradeable Windows 10 PC like me, here's what you're missing with advanced Copilot AI.