Context-Aware Dictation Stack
Adapt dictated text to the destination, personal style, and corrections over time.
- Difficulty
- Moderate
- Time to result
- ~weeks to results
- Steps
- 6
- Confidence
- 96%
The Context-Aware Dictation Stack improves voice input through three layers. First, it recognizes the active destination so an email receives greetings, signatures, lists, and email-like prose while a Slack message receives a more conversational structure. Second, it applies personal preferences such as punctuation, tone, spelling, and formatting. Third, it learns from keyboard corrections and explicit rewrite requests, reducing repeated cleanup over time. Because the tool operates at the system level, the transformed text appears inside the application already being used rather than requiring copying between a recorder, an AI editor, and the final destination. The result is not merely a transcript but context-specific writing that increasingly resembles the user.
Origin
Alan demonstrates an email whose greeting, signature, list, punctuation, and tone are shaped by application context and learned preferences. Extracted from Marketing Against The Grain.
Core principles
- 01The same thought needs different packaging in different applications.
- 02Destination context should determine structure and tone.
- 03Personal writing preferences should persist across sessions.
- 04Corrections are training signals for future output.
- 05Keep the user in the active application rather than forcing context switching.
How to run it
- 1
Work in the destination
Open the email client, messaging app, document, or AI interface where the final text belongs.
Pro tip Use a system-level dictation tool that can insert text into the active application.
Watch out Check the tool's privacy and accessibility permissions before granting broad access.
- 2
Dictate the intended message
Speak the content naturally, including verbal cues for lists, recipients, tone, or purpose when useful.
Pro tip Supply the information, not painstaking formatting commands.
Watch out The application context cannot infer facts or intentions you never express.
- 3
Apply destination formatting
Let the system adapt greetings, signatures, paragraphing, lists, and conversational style to the active app.
Pro tip Compare the same message across email and chat to calibrate expected differences.
Watch out Automatic context detection may occasionally select the wrong style.
- 4
Correct or restyle by voice
Highlight unsatisfactory text and request a change such as making it more casual or formal.
Pro tip State one clear stylistic direction at a time.
Watch out A rewrite can alter meaning, so inspect more than tone.
- 5
Teach persistent preferences
Make consistent corrections and add specialist names or niche terms so future outputs better match the user.
Pro tip Correct recurring punctuation and spelling choices instead of silently accepting them.
Watch out Inconsistent corrections create ambiguous preference signals.
- 6
Verify before delivery
Review the adapted message for fidelity, personal voice, and suitability for the recipient.
Pro tip Pay special attention to names and relationship-specific language.
Watch out Personalization should assist judgment, not replace it.
In the wild
Alan dictates a message to John, verbally enumerates three benefits, and closes with his name. The system formats the greeting and signature, turns the benefits into a list, removes filler, and adds the exclamation-mark style it has learned from his prior writing.
→ A sendable email appears directly in Gmail without manual restructuring.
Alan highlights the generated email and asks Willow to make it more casual. The text is rewritten in place instead of being copied into a separate chatbot and then copied back.
→ The tone changes while the user remains in the original workflow.
Common mistakes
Using one style everywhere
An email, Slack message, and personal text may express the same idea but require different packaging. Ignoring destination context creates extra editing work.
Failing to correct repeated errors
If a system learns from edits, leaving recurring spelling or punctuation mistakes uncorrected prevents useful personalization.
Trusting personalization blindly
A learned style can still be inappropriate for a particular recipient or high-stakes message. Review the output in context.
Is it for you?
Best for
It is best for people who communicate across several applications and need each message to retain an appropriate format and recognizable personal style.
Not ideal for
It is not ideal when a device cannot safely access application context or when every output requires a rigid externally controlled template.
From the transcript
“So it's context aware. So it's able to understand that if for an email, it should be style like an email. If I were in…”
“If I don't like it, it learns over time so that it doesn't make that mistake again.”
“It replaces a lot of that context switching between apps.”
From the episode
Everyone’s Using AI Wrong – This Is the Real Unlock