Blog
Engineering notes on AI agents, automation, and the infrastructure behind them.
Self-Hosted Voice AI: Why GDPR Is the Wrong Test (NIS2 Is the Real One)
A GDPR tick isn't a NIS2 test. What you really need to verify with hosted voice AI vendors before NIS2 puts the board on the hook personally.
Splitting Grounding from Reasoning in Browser-Agent Stacks
Browser-agent stacks bundle grounding and reasoning. A local 2B parser splits them, beats GPT-4o on ScreenSpot-v2 by 2.5x, costs $4 to train.
Context Engineering Is Just File Naming
Context engineering sounds new. It is the file-naming hygiene developers always had, load-bearing now because LLMs read what you point them at.
Your AI Workflow Doesn't Need Better Prompts. It Needs Less AI.
Prompting is discovery. Skills are repetition. Gates are how AI workflows become reliable.
Agentic Knowledge Base — Karpathy's LLM wiki, with adapters
A framework that turns whatever task or note app you use into a Karpathy-style LLM wiki. Pluggable adapters, parallel retrieval with RRF.
What Anthropic's April 23 Postmortem Reveals About Your Agent Harness
Three bugs over two months, one usage-limit reset for every Pro subscriber. The postmortem reads like a free audit checklist for any production agent harness.
Voice AI in Production: From RunPod to Hosted Kubernetes
One pod serves one user at a time. Production serves thousands. Here's what that gap actually costs, and why voice AI companies keep asking for hosted Kubernetes.
95% of PII Redaction Doesn't Need an LLM. The Other 5% Does.
When to use deterministic masking and when a fine-tuned LLM earns its compute on SAP production data copies. A hybrid architecture.
AI Automation for Freelancers: 8 Workflows I Run Daily
Ai automation for freelancers, broken down into 8 concrete workflows I run on a €15/mo VPS. Proposals, invoices, briefings, CRM, content, all with Claude. Download the free AI Automation Checklist.
Telegram Bot with Claude API: 30-Line Bash Daemon
Build a Telegram bot powered by Claude in 30 lines of bash. Long-polling daemon, MCP tool access, attachments, allowlisting. Runs on a VPS.
What llama.cpp's Pace Tells You About On-Prem LLM Readiness
Your team asked for GPU budget for self-hosted inference. You said not yet. The tooling moved, the org didn't, and the delay is costing you leverage you don't know you're losing.
MCP Servers Explained: What Model Context Protocol Does
MCP servers explained without the marketing. What Model Context Protocol is, how the architecture works, and when to write your own server. Download the free AI Automation Checklist.