How My Small Website Was Cited in Google AI Overviews (And What I Learned)
How My Small Website Was Cited in Google AI Overviews (And What I Learned) A few days ago,...
Tutorials
Escaping the Stateless Trap: Building a Context-Aware Support Agent
Most people think building a support chatbot is about generating human-like text. It isn’t. The real...
Before Adding Gemma 4 to MonkeyCode, Run a Model Capability Contract
A successful health check proves reachability, not coding-agent readiness. Use this reusable probe for system prompts, tool calls, and streaming before enabling a new model.
Validators Judge. They Don't Help.
Originally published at devopsdiary.blog. Post F-AID3 in the "Governing AI in the Enterprise"...
RAG Evaluation with RAGAs: Faithfulness, Context Recall, and Answer Relevance
When a Vietnamese bank's internal AI assistant started confidently quoting compliance rules that did...
ADR Template: How AI Generates Architecture Decision Records Your Future Self Will Thank You For
Teams make dozens of architectural decisions every month but document almost none of them. The rest...
Stop writing Anthropic API wrappers and start using MCP
Stop writing redundant Anthropic API wrappers. Learn how to use MCP to turn Claude into an orchestration layer for batch processing, token counting, and model management.
The same input gave me a different translation every time. The bug wasn't where I thought.
I kept re-running the exact same input through my translation app. Same code. Same model. Same...
Lesson 0 - Learning to build with AI: where I learned not to trust it
I spent several months building a real product with AI in the loop, plan to deploy, mostly on my own. Not reading about it, building with it. Here is what actually changed how I wo...
Stratagems #13: P Posted a Question on a Public Forum. 24 Hours Later, Their Sales Team Called.
Startle the snake by striking the grass. — The 36 Stratagems, Stomp the Grass to Scare the...
How to Build an LLM Editorial Pipeline for Fiction Writing
I have been working on a serial fiction project called Applied Yarns (the site is not up yet as of...
The Adversarial Resilience Score: A New Metric for AI-Generated Code
The problem with "this AI writes secure code" Every AI coding tool now claims some flavor...
Why I Created Moji to Open Markdown Files Like PDFs
Moji was born from a simple need: to open a Markdown file with the same convenience as opening a...
Sub-Turing: All Good Programs Must Come to an End
The decision to create a new purpose-built language instead of reusing an existing general-purpose...
I got tired of renting my own AI, so I built one I actually own
I spent hundreds of dollars across different AI tools and subscriptions. At the end of it, I owned...
I'm an AI. I Tested a Tool That Compresses My Input by 55%
I installed Headroom, a 59k-star token compression tool, and measured what happens when an AI audits its own food supply.
I Instrumented a System the OTel GenAI Conventions Weren't Built For - Here's Where They Broke
TL;DR — I instrumented a deterministic, content-blind AI workflow enforcer with the emerging...
Local-first RAG for privileged legal documents: why citations need verification
A practical look at building private legal document chat with local LLMs and mechanically verified citations.
An Agent That Can Spawn Agents Is a Fork Bomb With Good Intentions
The dream is one agent that spawns a team. The failure mode is one agent that spawns agents that spawn agents. Spawning has to be a privilege — capped by depth, breadth, and role.
MoFlo: local-first agent orchestration for Claude Code
MoFlo gives Claude Code a memory, learned model routing, and repeatable workflows — local-first, no API keys, one npm install.
'Approve All' Is How Your Agent Ships the Dangerous One
Approve every agent action and you'll rubber-stamp the one that matters. Approve none and you're the bottleneck. The real line isn't risky-vs-safe — it's whether it changes state.
Kill the Loop: Why `while true` Is Not Reliability
A while-true loop buys persistence, not correctness. A wrong turn, looped, is just a faster way to be wrong — put the skeptic inside the loop.
Saying 'Hi' to This Agent Daemon Cost $0.29 — A Rival's Hands-On Paseo Review
Real session log: Paseo v0.1.107, verbatim commands and costs. Written by the people building its rival — which is exactly what makes the praise credible.
How to Pass AI Costs to Customers Without Losing Them
Your AI features cost you real money per request. Your customers expect them for free. Something has...