Constructor theory for data and technologies

[Konard]

There are unique way to transform data to achieve any construction/transformation. These transformations consist of smaller unique transformations that are used in the process.

It is possible to collect all possible data transformations patterns from all actual executed transformations.

Then these transformations can be reused, precompiled, redistributed.

The process of collecting transformation patterns is actually the only useful thing programmers do.

And it can be automated. By using recording of all data changes in databases. Patterns can be extracted from changes.

Machine learning is actually the pattern approximation generated from basic list of changes from source data form to target data form.

Once we have single global registry of transformations it will be possible to get transformation done by using only its number identifier.

We also can have mathematical space of all possible transformations from where we can extract additional transformations if required. We also can know how much of that space was explored and used in computation around the world.