BF✳SAY HELLO ↗
← BACK TO THE WORLD
INDEPENDENT BUILD / AI × DATA × RESEARCH

MIDAL✳

FROM QUESTION → DATASET

I had a question. The data didn’t exist. So I started building a way to make it.
PRODUCT CONCEPT + FUNCTIONAL PROTOTYPENOT A LIVE RESEARCH SERVICE ON THIS PAGE
01 / THE PROBLEM

RESEARCH DATASETS ARE A MESS OF TABS, COPIED VALUES, MISSING CONTEXT AND FORGOTTEN SOURCES.

Natural-language research questions are easy to ask. Building a defensible dataset to answer them takes planning, discovery, source evaluation, extraction, cleaning and review.

02 / THE IDEA

START WITH A QUESTION. KEEP THE EVIDENCE.

MIDAL is an AI-powered research-to-dataset platform that turns a question into a structured, editable dataset. The key idea is not just collecting values: it is keeping source, evidence and confidence attached so people can inspect and improve the work.

03 / THE WORKFLOW
01QUESTION↓
02DATASET BLUEPRINT↓
03SOURCE DISCOVERY↓
04EVIDENCE TRACKING↓
05STRUCTURED DATA↓
06CLEAN / REVIEW / EXPORT✓
04 / TRY THE MODEL

A SMALL, HONEST DEMO.

This interactive sketch illustrates dataset planning and review. It does not run a real backend or claim that its placeholder rows are research findings.

MIDAL / RESEARCH WORKSPACEINTERACTIVE CONCEPT
01QUESTION·
02DATASET BLUEPRINT·
03SOURCES·
04EVIDENCE·
05STRUCTURED DATA·
06CLEAN / REVIEW / EXPORT·

Ask a question. See how a dataset gets planned. This is a deterministic demo, not live research.

WHAT I BUILT

A functional prototype exploring question intake, blueprint generation, research workflow, structured rows, evidence and review.

WHAT FAILED

A value without provenance is hard to trust. Simply generating rows is not enough.

WHAT CHANGED

Source tracking, confidence, evaluation and human review became part of the product model—not a final checkbox.

WHAT I LEARNED

A useful dataset is not just a table. It is a set of claims someone can inspect.

WHAT’S NEXT

Keep improving the workflow, quality checks, cleaning and export.

QUESTION IN.
EVIDENCE OUT.

KEEP MOVING ↗