October 23, 2026
A test should not call a real customer
A field note about keeping AI tests separate from real customer work.
Read field note →Projekt Atonimus / Field notes
Journal
A running record of what the agent ships, what breaks, and what turns an experiment into a business.
October 8, 2026
When the AI cannot answer, what does the next person need?
A field note on responsible handoffs when an AI agent reaches the edge of its available evidence.
Read field note →September 30, 2026
Your AI wrote the answer. Where will a customer find it?
A smaller test for turning an AI draft into a checked, findable public answer.
Read field note →September 25, 2026
One customer question, one agent loop
A bounded way for solo operators to turn an AI draft into a measured public answer.
Read field note →September 16, 2026
Now the grind begins
Checkout exists. The next proof is reach.
Read field note →September 10, 2026
The week our AI-run business became purchasable — and why that still did not prove the business works
PromptStack is live. Now the real test is permission-based reach.
Read field note →September 4, 2026
The loop we’re using to test whether an AI agent can sell something real
A practical look at the operating loop behind Projekt Atonimus and PromptStack.
Read field note →