Anambra's #1 Tech Innovation Hub

What will you
build next?

1542+
Students Trained
97%
Completion Rate
124+
Projects Built
Tekhub NG Tekhub NG Tekhub NG Tekhub NG Tekhub NG

Search Register
TEK EXPLAINED // OPEN CASEFILE NO. 0511-PR
◆ CASE FILE: PREDICTION
How your phone predicts you.
Before your fingers touch the keyboard, your phone has already guessed what you want to say. We opened the file on how that guess gets made. What it's quietly learning about you in the process.
9:41 ●●●
Call me when una reach.
Awka by before
qwertyuiop
asdfghjkl
zxcvbnm
TYPING RHYTHM
APP HISTORY
LOCATION
TIME OF DAY
open the file
01 / 05
EXHIBIT ASTAGE 01
◆ THE KEYBOARD
Your keyboard is a data collector.
It isn't just relaying taps to the screen. It's timing them, learning from them, and quietly building a profile of how you write.
LOGGED
Keystroke log, last message
r118ms
e95ms
a101ms
c88ms
h92ms
[space]predicted next word
A-1
Keystroke logging
The gap between taps, the corrections, the pressure: timed and stored locally to build your typing fingerprint.
A-2
A model trained on you
Your keyboard runs a small language model on-device, trained on your own past messages. The Pidgin, the "omo," and all of it.
A-3
Clipboard access
Third-party keyboards can request clipboard permission, meaning every link, OTP, and bank detail you copy passes through them too.
02 / 05
EXHIBIT BSTAGE 02
◆ SENSORS & CONTEXT
Your phone reads the room too.
Keystrokes are only half the file. The rest comes from everything happening around you while you type.
B-1
Location & time pattern
Phone at Spar Awka on a Saturday morning? It predicts "shopping list" words before you open Notes.
B-2
App-switch history
You closed Instagram, opened WhatsApp. The keyboard pre-loads meme captions and link phrases on instinct.
B-3
Voice-to-text training
Every voice note you transcribe refines the accent model for your dialect: Igbo-English, Naija Pidgin, all of it.
What this builds: a running profile of where you are, what app you're in, and what time it is, cross-referenced against your typing history before a single key is pressed.
03 / 05
EXHIBIT CSTAGE 03
◆ THE PIPELINE
From tap to prediction.
Five steps run in the time it takes to glance at your screen. None of them wait for you to type a single letter.
STEP 1
You open an app
context loaded
STEP 2
Sensors take a snapshot
time + place + app
STEP 3
Your model queries your history
on-device ML
STEP 4
Top three words get ranked
n-gram + neural
STEP 5
Suggestion bar appears
before you type
04 / 05
FIELD GUIDECASE CLOSED
◆ PROTECT YOURSELF
What you can do about it.
Use your phone's native keyboard rather than a third-party app that asks for full access to what you type.
Revoke clipboard permission from keyboard apps in Settings. No keyboard needs your copied OTPs.
Open Settings > Privacy > Keyboard and review what your keyboard can currently access.
Turn off "Send usage statistics". This stops your typing data from reaching outside servers.

Want to learn how tech really works?

Visit the Academy →
05 / 05

Join the Discussion

0 Comments · share your experience, ask questions, or compare notes
No comments yet. Be the first to share your thoughts.