01GUIDE · 8 MIN
How to spot an AI automation worth building
Volume, variability, return, control and a concrete way to decide whether to build.
Read resource↗OPEN KNOWLEDGE / 001—010
A library for building with judgment.
Guides and lab notes you can use before hiring me. Decisions, tests and architectures explained from practical work.
Subscribe via RSS↗DOWNLOADABLE RESOURCE · 5 PAGES
A checklist for moving beyond “it seems to work”.
Take it into a real session and document scope, experience, security, AI, games, evidence and the release decision.
Download in English↓CAPABILITY BRIEF · 4 PAGES
What I build, teach and validate, in four pages.
A shareable summary of consulting, training, Beta Lab, the laboratory and public evidence. No rates, inflated promises or sensitive names.
Download brief↓Tool / Process cost
Example for the process calculator
Before automating, do the maths.
Documents per week40
Minutes per document6
Current review workload240 min / week
40 × 6 min = 4 h
Estimated current workload. Deduct review, maintenance and exceptions when calculating savings.
01Measure the actual work
02Test a representative sample
03Compare time and errors
Sample data
02CHECKLIST · 10 MIN
What an AI-built product must test
Quality, attacks, tools, fallback, cost and regression before launch.
Read resource↗03LAB NOTE · 9 MIN
MCP, skills and agent teams without the hype
What each piece solves and how to design autonomy that produces evidence.
Read resource↗04GUIDE · 12 MIN
How to design an agent team that knows when it is done
Managers, handoffs, work contracts, evaluation and four verifiable exit states.
Read resource↗05LAB / 001 · 14 MIN
RogueBall: audited Godot artifact
Code, configuration, hashes, verified mechanics and honest agent-evidence limits.
Read resource↗06GUIDE + TOOL · 7 MIN
How to write an AI project brief
Turn outcome, process, data, risk and success criteria into a copyable document.
Read resource↗07GUIDE + CALCULATOR · 7 MIN
What a manual process actually costs
An explicit baseline for time, volume, people and rework, without promising future savings.
Read resource↗08GUIDE + PLANNER · 9 MIN
How to run the Google Play closed test
Separate the 12-testers-for-14-days threshold from the evidence, builds and corrections that explain the test.
Read resource↗09GUIDE + SELECTOR · 10 MIN
How to choose a Steam Playtest route
Limited or open Playtest, keys, internal access or a beta branch according to the game, audience and decision.
Read resource↗010GUIDE + SELECTOR · 10 MIN
How to run a TestFlight beta
Internal groups, external testers and public links separated by review, identity, capacity and evidence.
Read resource↗