Emissions Workflow MVP
CO2Calc
LiveA trilingual workflow automation for tour CO2 tracking — built in two days, funding-ready, and preserved as a frozen teaching snapshot.
Core Framework
Use AI without losing responsibility for what gets built.
AI can speed up product work. It can also speed up confusion, brittle code, and undocumented decisions. Judgment Before Automation is a framework for keeping architecture, review, and accountability in human hands while still using AI fully.
The Practical Problem
AI can accelerate delivery and still make systems harder to trust.
The failure mode is familiar: teams move fast, but the repo gets messy, decisions become hard to trace, and cleanup gets pushed to whoever inherits the project. Speed is not the problem. Unstructured speed is.
What usually breaks
- Specs come after generation
- AI decisions go undocumented
- Technical debt appears early
- Responsibility becomes blurry
The issue is not AI use. It is unbounded AI use.
What JBA Means
JBA keeps human judgment in charge of the build.
In practice, that means humans decide the problem, the constraints, and the approval points before AI starts generating output. AI is used for execution, but inside a structure that can still be explained, reviewed, and handed off.
What changes in practice
- Define the problem before prompts
- Write boundaries before generation
- Review before anything ships
Who this helps
- Founders and product leads
- Designers moving closer to implementation
- Developers who want cleaner AI-assisted workflows
The method inside JBA is Human-in-the-Middle. The MVP Lab is where that method is taught in depth.
The Method Inside JBA
HITM gives humans authority before AI output exists.
Human-in-the-Loop usually means a person checks output after it appears. Human-in-the-Middle means the human defines the structure, the boundaries, and the approval points before generation begins, then still decides what ships.
Core sequence
- Research is human
- Structure is human
- AI executes stepwise
- Nothing ships without review
HITM vs Human-in-the-Loop
- Human-in-the-loop reviews after generation
- Human-in-the-middle shapes the process before generation
- HITM keeps architecture and approval human-owned
Applications
Where this becomes useful
JBA matters anywhere teams want AI speed without losing traceability, reviewability, or handoff quality.
- MVP development
- Product architecture
- Repository discipline
- Asset pipelines
- AI-assisted design workflows
- Funding-ready packaging
More Than Prompting
A systems approach, not a trick
Judgment Before Automation is a way of organizing decisions, constraints, and review. Prompts are tools inside that system. They are not the system.
Proof
These are real builds, not thought experiments.
Four projects, different domains, one repeatable method. Each case study documents the problem, the process, and what actually shipped.
Street Ritual Game
Spiegel des Universums
LiveA location-aware street game with phrase generator, photo pinboard, and n8n orchestration — built in one day after three days of failure without structure.
Editorial Design
PDF to Web — 400 Pages
In developmentA large editorial publication transferred to clean, maintainable HTML/CSS. The errors became the spec. The last eight chapters required zero corrections.
Game Engine
Hand-Drawn Map Quest System
Open-source release plannedA quest-and-movement engine for hand-drawn maps. Built in an afternoon as a proof of concept after a developer said it was impossible without vector graphics.
All case studies are written from the actual build process — not reconstructed after the fact.
Flagship Program
The Human-in-the-Middle MVP Lab is the primary applied format of this framework.
Build an MVP in 5 days without losing architectural control. The MVP Lab is the high-commitment, full-system experience under the JBA framework.
Relationship to Context
Context asks the questions. JBA builds the structures that respond.
For deeper analysis of AI discourse, systemic bias, and broader questions, see Context. JBA focuses on applied architectural responses and responsible implementation.
Positioning
Judgment Before Automation is not a trend response. It is an architectural stance.
AI will accelerate. Judgment must scale with it. This page introduces the framework; the HITM page shows the model and the MVP Lab applies it.