A model becomes useful only when it survives materials, constraints, people and unexpected reality.
Workshop is where uncertainty becomes visible. We build the smallest useful test, expose the boundary, observe what fails, and keep only what reality supports.
Assembly / Verification
The projection is not expected to remain visually perfect. Interfaces shift, constraints appear and implementations change. The question is whether the essential structure still survives.
Current Experiments
Can one capability move between human work, local intelligence and external AI without changing the surrounding system?
Human ↔ Local AI ↔ External API
Site Factory translation and research text analysis expose the same executor-selection problem in two different domains.
Can machine-assisted knowledge change without bypassing human judgment or corrupting canonical knowledge?
Draft → Match → Conflict → Human Verdict → Execution
Archivarius verifies a controlled decision pipeline before any mutation reaches canonical knowledge.
How much visual freedom can a site expose without dissolving template consistency?
Defaults → Configuration → Controlled Override
Site Factory separates content, design tokens, site-level settings and allowed overrides into explicit responsibility layers.
Verification Logic
A Workshop experiment is useful only if its result can alter the model. Success, failure and inconclusive evidence are all legitimate outcomes.
State what we believe should survive contact with reality.
Make the uncertain responsibility, constraint or interface explicit.
Build the smallest useful experiment that can actually change the conclusion.
Observe what happened instead of defending the preferred architecture.
Validate, revise, archive or return the idea to the Lab.
Workshop → Works
When an idea survives its defined constraints, it can move toward production. If it does not, the result is still useful: the evidence returns to the Lab and changes what we think the system really is.
Continue to Works