Inctasoft Blog
Insights on agentic engineering, AI workflows, and building software that works.
Setpoint Evals — Slide Edition
Slide-format companion to the Setpoint Evals article. One big idea per slide. Renders as a vertical-scroll post or exports to a Reveal.js / PDF deck via Marp.
Setpoint Evals: Giving AI Coding Agents a Long Horizon
How shell-based end-to-end tests become long-distance acceptance criteria that let an AI coding agent work autonomously for hours — and a real codebase you can clone to try it yourself.
The Setpoint Problem: Why PID Loops for AI Agents Need More Than a Sensor
An AI coding agent with a perfect controller still goes to the wrong place if its setpoint is vague. Setpoint Evals are the setpoint definition language — and refusing to act without one is the next frontier of agentic engineering.
The PID Regulator for AI Agents: a Tiny Hook That Made My Coding Agent Self-Improving
How a small per-prompt context-injection hook closes a real feedback loop with your AI coding agent — and why thinking of it as a control system, not a prompt, changes everything.
The 2009 Profile That Wouldn't Die
There's a freelancer.com profile of mine still online from 2009, claiming 'for 2 years already I work successfully as a free lancer.' What that fossil says about the work I do now.
Hello, World
The Inctasoft blog — where we share what we learn building AI-powered products and agentic engineering systems.