RubilDownload for Mac

Quality status · Last run October 1, 2026 · Suite v3.1.0

Post-Edit Rate 0.13%

All four writing surfaces operational. 99.87% of dictations arrived ready to send with no edits at all.

Rubil is a voice dictation tool for Chrome, Mac, and Windows. You speak, and formatted text appears in whatever you are writing in: email structure in Gmail, a short message in Slack, clean prose in Notion. It formats your words and never writes new ones.

Latest run · October 1, 2026

99.87%ready to send

1 of 768 dictations needed an edit.

  • General writing0.34% Operational
  • Email0% Operational
  • Direct message0% Operational
  • AI chat prompt0% Operational

What this number is

Post-Edit Rate is the share of voice dictations that need at least one edit before they are fit to send. It scores send-readiness rather than transcription accuracy. Lower is better and zero is the goal.

Results by surface

Overall

41 runs, oldest to latest

1.09%across the window

41 runs agoLatest

  • General writing

    Operational

    Notes, documents, long-form

    0.34%Post-Edit Rate · 1 of 296 needed an edit

  • Email

    Operational

    Gmail, Outlook, Superhuman

    0%Post-Edit Rate · 0 of 160 needed an edit

  • Direct message

    Operational

    Slack, Teams, iMessage

    0%Post-Edit Rate · 0 of 236 needed an edit

  • AI chat prompt

    Operational

    Claude, ChatGPT, Gemini

    0%Post-Edit Rate · 0 of 76 needed an edit

  • Operational under 1%
  • Degraded 1–5%
  • Outage 5% and above

Run history

  1. October 1, 2026Suite v3.1.0Operational0.13%99.87% sent as written
  2. September 28, 2026Suite v3.1.0Degraded1.96%98.04% sent as written
  3. September 27, 2026Suite v3.1.0Degraded2.08%97.92% sent as written
  4. September 26, 2026Suite v3.1.0Degraded2.23%97.77% sent as written
  5. September 25, 2026Suite v3.1.0Degraded2.34%97.66% sent as written
  6. September 24, 2026Suite v3.1.0Degraded1.73%98.27% sent as written
Show all 41 runs
  1. September 23, 2026Suite v3.1.0Degraded1.56%98.44% sent as written
  2. September 22, 2026Suite v3.1.0Degraded1.74%98.26% sent as written
  3. September 21, 2026Suite v3.1.0Degraded1.82%98.18% sent as written
  4. September 20, 2026Suite v3.1.0Degraded2.12%97.88% sent as written
  5. September 19, 2026Suite v3.1.0Degraded2.61%97.39% sent as written
  6. September 18, 2026Suite v3.1.0Degraded1.70%98.30% sent as written
  7. September 16, 2026Suite v3.1.0Degraded2.49%97.51% sent as written
  8. September 15, 2026Suite v3.1.0Degraded1.06%98.94% sent as written
  9. September 14, 2026Suite v3.1.0Operational0.91%99.09% sent as written
  10. September 12, 2026Suite v3.1.0Degraded1.30%98.70% sent as written
  11. September 11, 2026Suite v3.1.0Degraded1.69%98.31% sent as written
  12. September 10, 2026Suite v3.1.0Operational0.65%99.35% sent as written
  13. September 9, 2026Suite v3.1.0Degraded1.05%98.95% sent as written
  14. September 8, 2026Suite v3.1.0Operational0.78%99.22% sent as written
  15. September 7, 2026Suite v3.1.0Operational0.93%99.07% sent as written
  16. September 6, 2026Suite v3.1.0Operational0.92%99.08% sent as written
  17. September 5, 2026Suite v3.1.0Operational0.92%99.08% sent as written
  18. September 4, 2026Suite v3.1.0Operational0.13%99.87% sent as written
  19. September 3, 2026Suite v3.1.0Operational0.13%99.87% sent as written
  20. September 2, 2026Suite v3.1.0Operational0%100.00% sent as written
  21. September 1, 2026Suite v3.1.0Operational0.26%99.74% sent as written
  22. August 31, 2026Suite v3.1.0Operational0.39%99.61% sent as written
  23. August 30, 2026Suite v3.1.0Operational0.44%99.56% sent as written
  24. August 29, 2026Suite v3.1.0Operational0.58%99.42% sent as written
  25. August 28, 2026Suite v3.1.0Operational0.87%99.13% sent as written
  26. August 27, 2026Suite v3.1.0Operational0.58%99.42% sent as written
  27. August 26, 2026Suite v3.1.0Operational0.14%99.86% sent as written
  28. August 24, 2026Suite v3.1.0Degraded1.02%98.98% sent as written
  29. August 23, 2026Suite v3.1.0Operational0.73%99.27% sent as written
  30. August 22, 2026Suite v3.1.0Operational0.44%99.56% sent as written
  31. August 20, 2026Suite v3.1.0Operational0.73%99.27% sent as written
  32. August 19, 2026Suite v3.1.0Operational0.14%99.86% sent as written
  33. August 15, 2026Suite v3.1.0Operational0.58%99.42% sent as written
  34. August 14, 2026Value-preservation guardOperational0.43%99.57% sent as written
  35. August 14, 2026Formatting model upgradeDegraded1.88%98.12% sent as written

What this measures

Post-Edit Rate is the percentage of voice dictations that need at least one edit before they are fit to send. The name comes from post-editing, the term translators use for the cleanup a human does on machine output. That cleanup is the real cost of dictation software, and this puts a number on how often you have to pay it. What is Post-Edit Rate? covers the full definition, the counting rules, and what the number does not tell you.

Every output is graded on four things. One edit in any category fails the whole message. Grading is deterministic; an independent AI judge reviews every output but cannot change a score.

  1. Content

    A fact missing, invented or changed.

  2. Voice

    Phrasing, hedging or intensity altered.

  3. Format

    The shape does not fit the surface.

  4. Mechanics

    Punctuation, capitalisation, sentence boundaries.

Each run sends the full scenario set through the live production pipeline several times over, because the failures that matter are intermittent and a single clean run hides them. Scenarios are written to be hard, so this is a stress test rather than a picture of an average day.

The bar is deliberately unforgiving in one direction. Rubil formats your words and never rewrites, generates or summarises them, so an output that reads more smoothly than what you said is a failure, not a bonus. Deleting a hedge like “I think”, softening “pissed off”, or tidying a dictated number all count against the score, because each one puts words in your mouth that you did not say.

Why each surface is scored separately

A dictation that is perfect for Slack is wrong for email. Averaging those into one figure hides the surface you write on every day, so each is graded against its own rules and reported on its own row.

  • Email

    Greeting on its own line, paragraph breaks at topic shifts, the closing the speaker dictated and no closing they did not. Names stay as spoken, so “Hi Sarah” never becomes “Hi, Sarah”.

  • Direct message

    Chat register: one or two compact blocks, greeting inline, no email scaffolding. Shorter in shape, never shorter in substance: if the email version makes four points, the message makes the same four.

  • AI chat prompt

    Structure is preserved rather than prettified. A dictated sequence renders as a numbered list, and an instruction stays an instruction: Rubil formats what you asked for, it does not answer it.

  • General writing

    Documents, notes and posts. Long dictation is broken into paragraphs at natural topic shifts, and values stay exactly as dictated, so a build number or a time is never quietly reformatted.

Why not Word Error Rate?

Word Error Rate is the standard for speech recognition and it answers a real question: how many words did the recognizer get wrong. It is the right metric for judging a recognizer and the wrong one for judging whether you can send the message. A dictation can score a perfect 0% Word Error Rate and still cost you a minute of cleanup.

Word Error RatePost-Edit Rate
CountsIndividual wordsWhole messages
AsksDid it hear me right?Can I send this as is?
JudgesThe recognizerThe finished output
Punctuation & structureUsually ignoredCounted as a failure

Common questions

  • Rubil is a voice dictation tool for Chrome, Mac, and Windows. You speak, and formatted text appears in whatever application you are writing in: email structure in Gmail, a short message in Slack, clean prose in Notion, across 20+ surfaces. It is speech-to-text, not text-to-speech, and it formats your own words rather than generating new ones. Rubil Pro costs $7 a month or $59 a year. Free is a 7-day Pro trial, then 1,000 words a week.

  • Post-Edit Rate (PER) is the percentage of voice dictations that need at least one edit before they are fit to send. It scores send-readiness rather than transcription accuracy. A dictation counts against the rate if a reasonable user would fix even a single character before sending. Lower is better and zero is the goal.

  • On the most recent run (October 1, 2026), Rubil’s Post-Edit Rate was 0.13%, meaning about 99 in 100 dictations arrived ready to send with no edits at all. The suite runs weekly and every result is posted here.

  • Word Error Rate counts how many individual words a speech recognizer got wrong against a reference transcript. Post-Edit Rate counts whole messages that needed human intervention before sending. A dictation can score a perfect 0% Word Error Rate and still fail on Post-Edit Rate: every word transcribed correctly, but no punctuation, no paragraph breaks, and filler left in, so the user still has to fix it. Word Error Rate measures the recognizer. Post-Edit Rate measures whether the user got to skip the cleanup.

  • Four things. Content: a fact is missing, invented, or changed. Voice: the speaker’s phrasing, hedging, or intensity was altered, so softening “kind of satisfying” to “satisfying” counts. Format: the shape does not fit the surface, such as a dictated list returned as a paragraph. Mechanics: punctuation, capitalisation, or sentence boundaries. One edit in any category fails the whole dictation, which is why the bar is high.

  • Weekly. Each run sends the full scenario set through the live production pipeline several times over, not a staging copy, so the number reflects what users receive. Results are posted here whether they improve or not.

  • Because a good message looks different in each. Email needs a greeting, paragraph breaks and a closing. A Slack message reads as one or two compact blocks with the greeting inline. An AI chat prompt keeps its structure and lists intact. Rubil detects the surface and applies the right shape, which is why the board reports each one separately rather than averaging them into a single figure.

  • No. Rubil formats the user’s own words and never rewrites, generates, or summarizes them. Hedges such as “I think” and “kind of” are preserved deliberately, because hedging is part of how a person speaks. Numbers dictated as digits are copied character-for-character, so an order number is never reformatted, and intensity is left alone.

Measured weekly · Last run October 1, 2026 · Suite v3.1.0 · Full history as JSON