01 / WORKSHOP
Build from a clear question.
Define the scope. Gather the sources. Develop an explanation or a change, with its assumptions and limits in view.
RECORDA proposal you can examine.
The ZAI method / Trust
Reality has
final authority.
A clear question. Evidence you can trace.
A decision someone is responsible for.
Our method connects research, engineering and careful review—so you can see how a conclusion was reached, and what remains unknown.
Claim≠Evidence≠Validated reality
01 / How the work moves
Three places.
One discipline.
WORKSHOP → AUDIT → VAULT is our way of moving from an idea to a documented decision. A missing piece of evidence can pause the work at any stage.
02 / AUDIT
Give the work a challenge.
Separate AI reviewers examine the exact version. Reproduce checks where possible, test failure cases and keep unresolved questions visible.
RECORDEvidence for a human decision.
03 / VAULT
Keep the decision traceable.
Retain the reviewed version, supporting evidence, limitations and authorization together. A change returns through review.
RECORDA record the next person can follow.
01Name the source
Identify the record, its date and what it actually reports. Keep that source attached to the finding. A production record alone does not establish gas available to a new project.
02Separate fact from interpretation
Reconcile identities and compare records. Make clear which statements are reported observations and which are engineering interpretations.
03Challenge the conclusion
Test assumptions and reproduce checks where possible. Record contradictions. Agreement between AI systems is not independent validation.
04Record the decision
A responsible person weighs the evidence and its limits. Confidence remains proportionate to the evidence. UNKNOWN is a valid result.
02 / What that means in practice
An honest unknown
is useful information.
A record can tell us that gas was produced. It cannot, on its own, tell us that gas is available for a new project. We keep that distinction visible.
03 / Intelligence & responsibility
Many tools.
One methodology.
Human authority.
Our Autonomous Engineering Layer brings people and AI tools into defined roles. The methodology governs the work as tools and models change.
Separate AI reviewers check exact, hash-identified versions. A hash identifies the bytes being reviewed; it does not prove that a conclusion is correct.
Model agreement ≠ independent validation. Separate AI checks are part of our process; they do not by themselves establish independent validation or independent human review. Evidence and reproducible checks support the Founder’s decision.
AI can make mistakes. ZAI requires evidence.
Final authorization remains with the Founder. Work returns through review when its scope or version changes.
AI tools support our internal development and public-source research.
Our customer-information rule
ZAI does not use AI tools to process customer-provided information.
This is ZAI’s handling rule for its own work. Email and voicemail providers process messages under their own terms; we do not claim an AI-use or training exclusion for those providers. Please send only public information through Reception.
04 / A method that can say stop
The correction is
part of the work.
Our development record includes pauses, changed assumptions and repeated checks. What matters is how the evidence changes the decision.
A stop was the right result.
When a release check could not confirm that an earlier test had ended safely, publication paused. The evidence was retained and the check corrected before further review.
The environment matters.
A successful test did not yet represent its intended operating conditions. The change was tested again under those conditions.
Small details have consequences.
Review caught a naming conflict before execution. The correction became part of the record, so the next change could build on it.
Examples from ZAI’s internal development process, described by the Founder. They illustrate working habits; asset qualification requires its own evidence.
Evidence first. Responsibility always.
Bring a question.
Build understanding.
Start with public information. We reply in writing.