LatestLLM
Prefix caching vs KV cache: what actually gets reused
Separate decode-time KV reuse from cross-request prefix caching, diagnose cache misses, and keep vLLM controls distinct from hosted provider APIs.
Read
Writing
Blog
Notes on cloud architecture, AI systems, and the details that only surface once something is running in production.
2026
Open WebUI
Open WebUI + FastAPI: fix model discovery and streaming
Microsoft Foundry
Foundry Agent Service: a production readiness checklist
LLM
KV cache memory sizing: GQA math and optimization
ego lite
ego (lite): a browser your coding agent is already logged into
LLM
Why LLM Inference Caches K and V but Never Q
Security
Why "Don't Store JWT in localStorage" — Auth History and the Cookie Return
GitHub Actions
Always Set persist-credentials: false in actions/checkout
Microsoft Foundry
Microsoft Foundry Agent Service Is Becoming a Real Cloud Runtime
CI/CD
Automating Azure App Service Deployments with GitHub Actions and Secure Secrets
OpenAI
GPT-5.4 Brings Stronger Coding, Tool Search, and Computer Use
Docker
Dockerizing a Complex App: Lightweight Images, Multi-Stage Builds, and Secrets
Next.js
Next.js + FastAPI: The Full-Stack Combination That Replaced Our Monolith
AI
Connecting Custom AI Agents to OpenWebUI: Auth, Latency, and API Design
Azure AI
SchemaForge: Extracting Structured Data from Unstructured Documents with Azure AI
2025
SQL Server
SQL Database Scheduled Backup to Azure Blob
AI Agents
Designing Agentic AI Workflows: Tool-Calling, Memory, and Multi-Modal Integration
AI
Building a Production RAG Pipeline with Azure AI Search and GPT-4
Azure
Building a Cloud Cost Calculator with the Azure Retail Prices API
Azure
Private Endpoints on Azure: The DNS Gotchas Nobody Warns You About
Azure
Securing Azure Virtual Desktop with Entra ID and Passwordless MFA
Azure
Monolith to Microservices on Azure: A Production Migration Playbook
Vercel
AI SDK 5 Pushes Type-Safe Agents Further Into the Frontend
OpenAI