Vedokrok Lab · open notebook
Finding a good page is one problem. Choosing a fitting move, seeing its limit and checking what happened are harder. This is where we show the questions we test and the work still missing.
An inspectable search check · 28 September 2026
Editors matched 37 task queries to expected pages in the 20 September 2026 edition (MHC-RPUB-20260920-75ad787a). Twelve unrelated queries checked that search could return no results.
Six pages ranked second or third. We revised the site after earlier misses and added one query to the same set. This repeat check can show whether those changes still work; it cannot tell us if readers find or use good advice. It says nothing about Google rankings or AI citations.
One real evidence trail
This example comes from the selected public edition. The source supports a specific description; it does not prove that using the idea will improve a decision.
The claim
Confirmation bias includes seeking or interpreting evidence in ways that favor an existing belief or hypothesis.
The source and its role
Role: supports. Supports only the scoped description; does not test the corpus procedure.
Where to look: Publisher abstract and review description
Open the sourceThe boundary
Keeping the same conclusion after strong evidence is not confirmation bias. The label describes a pattern in evidence handling, not disagreement itself.
Take it into a real decision
What result would make you weaken the current explanation? Have you looked for that result with the same effort and evidence standard you used for supporting evidence?
Item MHC-U-9C0A00000001 · version 0.2.0 · MHC-RPUB-20261011-9e4b62c5Questions on the bench
Our editorial search probes start with ordinary task phrases. Each names an intended page; separate nonsense queries check whether search knows when to return nothing.
Check: Retrieved page and rank, including misses and false positives.
Try the searchA worked case now includes a small, local checkpoint you can copy. We have not measured whether it helps people resume faster or with fewer mistakes.
Check: Return time and reconstruction errors against a baseline.
Inspect the caseThe public agent endpoint can retrieve a released item with its version and source trail. A fluent response alone would not prove that the person made a better decision.
Check: Correct retrieval, stated limit and an observable task outcome.
See the current agent accessShow the seams
The site currently reads selected release MHC-RPUB-20261011-9e4b62c5. The MHC corpus maintains the underlying knowledge and evidence; Vedokrok checks retrieval of its approved public projection and plans to study how people use it. We have no published participant outcome study, independent efficacy claim or verified search ranking to present.
Work with the lab
A useful proposal names the task, the comparison, what can be observed and what result would count against the idea.