I did not start Vahtian because research needed another dashboard. I started it because I kept meeting decisions that mattered, but were difficult to reconstruct later.

A sampling choice lived in a meeting note. A codebook changed without a reason beside it. A claim sounded right but had drifted beyond its source. An AI suggestion entered the work, while the human decision that followed was left implicit.

Can someone else see why this decision was made?

The gap I kept meeting

Most research problems are not dramatic. They accumulate through ordinary work:

  • why recruitment stopped when it did;
  • why one interpretation was kept and another rejected;
  • whether a citation really supports the sentence beside it;
  • what an AI system suggested, and what a person actually accepted.

None of these decisions is suspicious on its own. The problem is losing the path between the input, the judgment, and the final claim. Months later, the work may still look polished while the reasoning underneath has become hard to inspect.

Software should make the work more inspectable

That is the job I give Vahtian. A tool can structure a question, run a check, compare versions, or flag a mismatch. It should also leave enough of a record for someone to understand what happened.

This matters even more when AI is involved. A confident answer is not the same as a defensible decision. The researcher still has to decide what counts as evidence, whether a method fits, and what belongs in the record.

Why the core stays open

Methods used to check research should themselves be open to inspection. That is why the methods engines are published as open-source packages:

  • RecoverLite asks whether a planned design can recover what it claims it will estimate.
  • EpiNet makes leakage, calibration, and uncertainty visible in graph-based analyses.
  • CiteVahti checks whether a source supports the claim written beside it.
  • AssessLite stress-tests the assumptions behind a causal result.

Open source does not make a method correct. It makes inspection, testing, and disagreement possible. That is the more useful promise.

What belongs here

This blog is where I write through the problems behind the tools: research design, qualitative work, local AI, scientific writing, and the awkward decisions that do not fit neatly into a methods section.

I will show what I tried, what held up, and where the approach still falls short. The point is not to make research look frictionless. It is to make the reasoning easier to follow.

Better research is not only more efficient. It is easier to explain.

That is the standard Vahtian is built around.

Heidi