Writing
- AI Security · Zero Trust (opens on Medium in a new tab) Summary on this site: Zero-Trust Security for Autonomous AI Agents →
Zero-Trust Security for Autonomous AI Agents
Why system prompts won't save your MLOps pipeline, and how mathematical formal verification will. The Agent Skills Security Oracle combines zero-trust ABAC, real-time inference steering and Z3 proofs to stop agents making unauthorized tool calls. - GKE Autopilot · Load Testing (opens on Medium in a new tab) Summary on this site: Solving the LLM Load Testing Paradox: A Reference Architecture on GKE Autopilot →
Solving the LLM Load Testing Paradox: A Reference Architecture on GKE Autopilot
Hardening AI agents for production with GKE Autopilot, Parlant and OpenTelemetry, so you can load test your infrastructure without burning your LLM budget on real model calls. - NanoChat · LLM (opens on Medium in a new tab) Summary on this site: Reproducing Karpathy's NanoChat on a Single GPU — Step by Step with AI Tools →
Reproducing Karpathy's NanoChat on a Single GPU — Step by Step with AI Tools
A fully runnable notebook version of Karpathy's NanoChat, reproduced on a single GPU using DeepWiki, Gemini and Modal. - Legacy Modernization · Enterprise AI (opens on Medium in a new tab) Summary on this site: AI is Ready. Your Legacy Database Isn't. Let's Fix That. →
AI is Ready. Your Legacy Database Isn't. Let's Fix That.
Project Concord is a high-performance, TypeScript-based MCP server designed to help AI systems work smoothly with older business databases by turning messy and unclear data into clean, reliable information. - Mcp Server · MCP (opens on Medium in a new tab) Summary on this site: Building Your First No-Code MCP Server: The Fabric Integration Story →
Building Your First No-Code MCP Server: The Fabric Integration Story
How I created a Model Context Protocol (MCP) server for Fabric REST API without writing a single line of code - AI · Content Generation (opens on Medium in a new tab) Summary on this site: Introducing Talking Book →
Introducing Talking Book
I created Talking Book to make reading more engaging and accessible, especially for my 9-year-old son with dyslexia. - Local AI · Private AI (opens on Medium in a new tab) Summary on this site: Beyond the Cloud: How I Built My Own AI Server (and Why) →
Beyond the Cloud: How I Built My Own AI Server (and Why)
I built my own AI server using an Orange Pi 5 Max, Ubuntu 24.04, Docker, and LocalAI, enabling me to run various open-source models locally. - Open WebUI · Private AI (opens on Medium in a new tab) Summary on this site: Would you pay $1/month to Own Your AI Data? →
Would you pay $1/month to Own Your AI Data?
Our AI conversations are valuable assets that shouldn't be surrendered to Big Tech without control or ownership. - FastRTC · WebRTC (opens on Medium in a new tab) Summary on this site: Talk To AI with FastRTC: Real-Time Voice Conversations with AI →
Talk To AI with FastRTC: Real-Time Voice Conversations with AI
Talk To AI turns text chat into a voice-driven experience using real-time speech-to-text, LLM generation and text-to-speech built on FastRTC. - Embeddings · OpenAI API (opens on Medium in a new tab) Summary on this site: Unleashing the Power of Open Source Text Embeddings with open-text-embeddings →
Unleashing the Power of Open Source Text Embeddings with open-text-embeddings
open-text-embeddings is an OpenAI API-compatible embeddings endpoint that supports open source sentence-transformer models and other embedding models.