Short answer: open EmEditor, click where the note should go, press Ctrl+Alt+R, speak the finding, press Ctrl+Alt+R again. AICHE drops clean text at the cursor. The 12 GB log keeps scrolling.
The Problem
EmEditor is the editor people reach for when other editors choke. 16 TB file ceiling. 1.099 billion lines. A 10 GB file opens in about a second. You use it because you're staring at things normal tools refuse to load.
The friction isn't reading the data. EmEditor handles that. The friction is writing about what you find. You spot a 503 spike in a filtered log, you decode what cust_seg_v2_adj actually means in column F, you finally get a regex right after twenty attempts. Then you stop, switch hands to the keyboard, and type a paragraph explaining it while your eyes are still tracking the data. By the time you finish typing, the mental thread is gone.
Speaking is 150 WPM. Typing is 40 WPM. More importantly, speaking doesn't require you to look away from the row you just found.
What Changes
You keep one hand on the mouse, scrolling and filtering. When you find something worth recording, click where the note should land (inline, in a markdown sidecar, in a comment block, in a notes tab), press Ctrl+Alt+R, talk through what you're seeing, press the hotkey again. AICHE transcribes, cleans the filler, and inserts at the cursor. You keep scrolling.
The file's size is irrelevant to AICHE. It inserts a few hundred characters at a cursor position. EmEditor's buffer handling does the rest, the same way it handles any keystroke in a 50 GB file.
How It Works
- Open EmEditor and load your file. Log, CSV, TSV, source, binary, whatever.
- Click into the spot where the annotation belongs. That can be the data file itself, a second tab, or a split pane.
- Press Ctrl+Alt+R.
- Speak the observation. Full sentences, the way you'd explain it to a colleague.
- Press Ctrl+Alt+R again. AICHE inserts cleaned text.
- Keep working. The hotkey is global, so it works the same in a normal
.txtfile as in CSV mode or the Large File Controller view.
First Workflow: CSV, Regex, Macro Notes
Start here when the file is too large to re-scan from memory.
CSV mode, while building a data dictionary
EmEditor's CSV mode shows delimited files as a real table with a frozen header row, Filter in Column, Pivot Table, Join CSV Tables, and Flash Fill. When you're working a CSV export with column names like cust_seg_v2_adj, acq_chnl_grp_id, mrr_norm_q3, you're constantly figuring out what each column means by scanning sample values and checking ranges.
Open a second tab next to the data and use it as a data dictionary. When you decode a column, click into the dictionary tab, press Ctrl+Alt+R, and describe it in plain language. By the end of the session you have a column-by-column reference, written while the discovery was fresh, not reconstructed from memory three days later.
Log investigation, while the trail is still in your head
A real incident review is a chain: filter for the error, see a cluster between two timestamps, cross-check with the deploy window, form a hypothesis. EmEditor's Filter Bar and regex search make the chain fast. The problem is that the chain only exists in your head while you're working it.
Pick a position in the filtered view (or open a sibling notes file), press Ctrl+Alt+R, and narrate the link you just made: what you filtered on, what you saw, what you think it means, what you're going to check next. When the on-call writeup is due tomorrow morning, the writeup is already there.
Regex you can read next month
EmEditor's regex engine handles lookaround, named captures, and Boost-style syntax. The patterns get complex fast. A regex you can decode the day you wrote it becomes unreadable in six weeks.
Drop a comment line above the pattern, press the hotkey, and describe what it matches and why. "Matches ISO 8601 timestamps where seconds precision is present. Group 1 is the date, group 2 is the time. Intentionally rejects entries without seconds because the upstream feed always includes them and absence means a parser error." Future you, or whoever inherits the macro, reads the explanation instead of reverse-engineering the regex.
Macros and Find/Replace, documented as you build them
EmEditor's macros are JavaScript or VBScript with a documented API, and Find/Replace can run regex against tens of millions of rows in one shot. Both are easy to fire and hard to remember. Before you run a global replace on a 200 million row file, talk through the pattern, the replacement, and the assumption you're making. The annotation is the rollback plan if the replace does something you didn't expect.
Large File Controller, with breadcrumbs
The Large File Controller opens a window into a file that's bigger than memory should reasonably hold. You navigate by byte offset, line number, or section. The breadcrumbs you leave behind matter more in this view than anywhere else, because there's no scrolling back through 1.099 billion lines on instinct. Drop a marker on the line, press Ctrl+Alt+R, dictate why this section matters. The marker takes you back to the line. The voice note tells you why the line mattered.
Practical EmEditor Setup
- Split window or second tab. EmEditor's split view lets you keep the data file open in one pane and a notes file in the other. Click into whichever pane you want the text to land in, then press the hotkey.
- Markers. After dictating an annotation in a huge file, set a marker on the line. EmEditor's marker list is your jump-back index for everything you flagged with voice.
- Custom vocabulary. If you work with internal service names, table names, or weird column abbreviations, add them to AICHE's custom vocabulary so they come back spelled the way your team writes them.
- Software Development profile (Pro). Tuned for code-shaped speech: API names, snake_case, kebab-case, file paths, regex tokens.
Common Questions
Q: Does the cursor insert really work inside CSV mode's table view?
A: Yes. CSV mode is still a text buffer underneath. Whichever cell or row your cursor is in, that's where AICHE inserts. For free-form notes you'll usually want a second tab or a sidecar file rather than typing into the structured data.
Q: What about the Large File Controller view? Will inserting text break the section window?
A: No. The Large File Controller is read/edit on a slice of the file. AICHE inserts at the cursor inside that slice exactly like any other keystroke. EmEditor handles the write-back.
Q: Will dictation slow EmEditor down on a 20 GB file?
A: No. AICHE runs out-of-process and hands EmEditor a single chunk of text at the cursor when you stop recording. EmEditor sees one paste, not a stream of keystrokes.
Q: I dictate technical terms, regex syntax, and SQL. Will it transcribe them correctly?
A: Turn on the Software Development profile and add your domain vocabulary (table names, library names, internal acronyms) to AICHE's custom vocabulary. Punctuation marks, brackets, and operators that you say out loud are transcribed as the symbol, not the word.
Q: Is there a Mac or Linux build of EmEditor I should know about?
A: No. EmEditor is Windows-only. AICHE runs on macOS, Windows, and Linux, but the EmEditor pairing only applies on Windows. The hotkey is Ctrl+Alt+R.
Q: Can I dictate into EmEditor in CSV cell-by-cell mode while moving with arrow keys?
A: Yes, but it's not what voice is good at. Voice shines for paragraph-scale notes, column descriptions, and regex explanations. For single-cell edits, the keyboard is faster.
Result: the gigabyte log gets reviewed with the analysis already attached. The million-row CSV gets a column dictionary built during the session that produced it. The regex that took an hour to write carries the sentence that explains it. The data and the documentation stop being separate jobs.
Try it now: open the largest file you've been putting off documenting, click anywhere in it, press Ctrl+Alt+R, and describe what you already know about it. Stop pretending you'll write the notes later.