BASALT · JOURNAL
Making a document process repeatable
A process performed by hand each time will eventually be performed differently, and in document work the differences are usually invisible until someone finds them.
Where variation creeps in
Steps done in a different order. Redacting after Bates numbering rather than before puts stamps in the wrong relationship to the marks.
A step skipped under pressure. Sanitisation is the usual casualty, because the redaction is the part that feels like the work.
Different settings. A compression preset chosen differently on Tuesday than on Monday produces two productions that do not match.
What to fix first
The order. Write down the correct sequence once and follow it: OCR, review and mark, redact, sanitise, number, verify, export. Most errors are ordering errors rather than execution errors.
What to automate
Anything with no judgement in it. Stripping metadata, flattening, applying Bates numbers, compressing, adding a watermark and running verification are all mechanical and identical every time.
Anything requiring judgement should stay manual. Deciding what to redact is the obvious case: a saved set of marks applied to a different document is not redaction, it is drawing rectangles in arbitrary places.
What a good automated run looks like
It shows you the plan before it starts, so you can see what will happen to which file.
It reports an outcome per file rather than a single completion message, because a batch that quietly skipped one document is worse than one that failed loudly.
It stops on a failure rather than continuing, since a document that silently failed to process is the one that gets produced.
It produces the same output for the same input. Reproducibility means a second person can run your process and confirm the result, which is the difference between a claim and a demonstration.
Verification belongs in the process
Not as a final manual step someone might skip, but as part of the run, positioned so that failure prevents output rather than warning about it. A check that produces a bad file plus a warning depends on somebody reading the warning.
Frequently asked questions
What should be automated in a document workflow?
Anything mechanical: stripping metadata, flattening, Bates numbering, compression, watermarking and verification. Anything requiring judgement, particularly deciding what to redact, should stay manual.
What is the right order for document production steps?
OCR, then review and mark, then redact, then sanitise, then number, then verify, then export. Most errors in document work are ordering errors rather than execution errors.
What should an automated batch run show?
The plan before it starts, an outcome for each file rather than a single completion message, and a stop on failure rather than a silent skip, since a document that quietly failed is the one that gets produced.
Why does reproducibility matter in document work?
Because it turns a claim into a demonstration. If the same input and settings always produce the same output, another person can run your process and confirm the result independently.
Doing it in Basalt
Basalt is a native macOS PDF toolkit: eighteen tools in one window covering merge, split, page organisation, compression, OCR, passwords, Bates numbering, forms, signing, watermarks and comparison. Every file is processed on your Mac, and the engine that opens documents holds no network entitlement at all, which macOS enforces at the code-signature level. Redaction destroys content rather than covering it, and an independent verifier proves the material is gone before a file is written. A one time $29 licence covers up to three Macs, free for the first 24 hours. basaltformac.com, or brew install --cask chipmunk1101/tap/basalt.
Redaction that proves itself
Basalt destroys the content you mark, then re-opens the file it wrote and proves the content is gone before it saves anything. Your documents never leave your Mac.