Change the format. Keep the meaning.

Start by deciding what must stay unchanged

Conversion is useful only if the result still represents the source. In a measurement, that means keeping the same quantity. In writing, it means preserving facts and intent. In a data table, it means retaining field names, identifiers, and the relationship between each row’s values. Make a short list of those invariants before changing the format.

For example, a customer code “0012” is not interchangeable with the number 12. A sentence saying a delivery “may arrive on Friday” does not mean it “will arrive on Friday”. And a US gallon is not an imperial gallon. A polished output can still be wrong in each of these cases.

Workflow 1: clean a pasted list

1. Paste the list into the text formatter.
2. Choose Clean spacing to collapse repeated spaces and trim line ends. Check that meaningful indentation was not removed.
3. Use the result as input, then remove duplicate lines. This compares exact text, so differently capitalised lines remain distinct.
4. Choose a case style only if it suits the destination. Title Case is convenient for labels but can damage names such as “iPhone” or acronyms such as “NASA”.
5. Copy the result and keep your source until the destination looks correct.

For a heading such as “ monthly REPORT ”, spacing cleanup gives “monthly REPORT”. Title case gives “Monthly Report”. The tool can make those mechanical changes predictably; it cannot decide whether “REPORT” was meant to be an acronym.

Workflow 2: review an AI revision

Use the DeepSeek writing editor when the problem is language rather than layout. Start with grammar mode if your wording is already close. Clarity and tone changes are broader and need a closer comparison. Submit only text you are comfortable sending to the AI provider.

Compare the revision sentence by sentence. First check names and quantities, then negations, dates, and words that express certainty. “We cannot confirm the date” must not become “We can confirm the date”. Last, read it aloud to check tone. The most fluent sentence is not necessarily the most faithful edit.

Workflow 3: move records from CSV to JSON

Suppose your table has columns code,name,quantity and a row 0012,"Notebook, blue",3. The comma inside the quoted name belongs to that field. The CSV ↔ JSON converter preserves the code as a string, the full name as one value, and the quantity as the string “3”. It deliberately does not infer numeric types.

Check the converted object before importing it. If a destination needs a number for quantity, apply that transformation explicitly in the destination. Do not convert every numeric-looking string at once: doing so would also strip the leading zeros from the code.

Check the reverse conversion too

When turning JSON into CSV, missing fields and null values become empty cells. Nested records are not supported because a flat CSV cannot preserve their structure without an additional convention. Spreadsheet-safe mode adds an apostrophe to formula-like cells. This protects a common spreadsheet workflow but changes the text; an exact machine-to-machine import may need that setting disabled after reviewing the source.

A small set of test records catches many conversion surprises.

Include a value like	Check for
0012	Leading zeros remain
Notebook, blue	The comma stays inside one field
She said “hello”	Quotes remain part of the text
An empty field	The row retains its column count
=1+1	It is treated as text when intended

Choose the smallest reliable tool

Use deterministic formatting for case and spacing, a parser for structured data, and AI suggestions for prose you can review. For physical quantities, use the unit converter and verify the unit definition before trusting the digits. A reversible workflow with an unchanged original makes mistakes easier to catch.