Labs
We test on ourselves first. This is what is running, what it is allowed to do, and what it has proved so far — which is nothing yet.
What is being tested
A system that reads Search Console and analytics data, ranks growth opportunities, drafts a change, routes it for approval, publishes it, and records what happened at 7, 14 and 28 days. The question is not whether it can generate content — that part is easy and mostly worthless. The question is whether it can produce growth measurable against a baseline, at a cost below the human time it replaces.
What it is allowed to do
Autonomy is staged. Today it is a copilot: it proposes, a person approves. Unattended action is unlocked only after the proposing stage has been shown to work, and only for changes that are cheap to reverse.
It may never, at any autonomy level:
- Change a price or any commercial term
- Change a factual claim or the evidence behind one
- Change a case study metric or baseline
- Set
noindex, alter a canonical tag, or create a redirect - Change a client name or publish one
- Change a permanent URL
Those prohibitions are enforced in code and checked against the built output, not asked for in a prompt. A system that can be talked out of a rule does not have that rule.
What is logged
Every action records what was done, why, on what, the hypothesis behind it, the model and prompt version used, the result expected and the result observed. Any figure this project ever publishes can be traced back to the run that produced it.
Current status
Baseline recorded 31 August 2026, from Google Search Console property for vianapulse.com, verified by the Owner. No results measured yet. Measurement windows are 7, 14, 28 days.
Follow it on the case study, which is updated as each measurement window closes.
If it does not work
That gets published too. An experiment that returns a negative result has still answered the question, and a lab that only ever reports success is not running experiments.