The AI in your pocket

Everything in this curriculum has quietly pictured a keyboard and a screen. For most people on earth the phone is not a secondary device — it is the only one.

This lesson is what changes when that is your situation.

Voice is the better input

On a phone, dictation usually beats typing for anything longer than a sentence, and people under-use it because it feels informal.

It is faster. It suits the long, context-rich prompts that E3 showed produce better answers — you will happily say four sentences of background you would never thumb-type. And rambling is fine: false starts, corrections mid-flow, “actually, no, what I mean is”. Models handle disfluent speech well, and E9·L3’s dictate-then-tidy pattern applies to everyone, not only people who find writing hard.

Two practical notes: check whether dictation runs on-device or in the cloud (L2’s question), and remember that you are speaking aloud — in an office, a train, or a shared home, that is a privacy decision about the people around you as much as the software.

The camera is an input

The phone’s real advantage over a laptop. Point it at:

A document, a form, a letter — text extraction and explanation.
A whiteboard after a meeting.
A product label, an ingredients list, a warning notice.
A screen showing an error message.
Something you cannot name and want identified.

E9·L4 covered this as an access tool with its failure modes; the same cautions hold for everyone. Digits are where it errs, and it will answer confidently about a blurred image unless you tell it to say when it cannot tell.

For anything with small print, ask for the exact text first, then the explanation. Checkable, quotable, and it exposes a misread before you act on it.

On-device changes the calculation

Modern phones run small models locally. Vendors advertise this, because it is a genuine advantage rather than marketing.

Works offline. No signal, no problem — on a train, in a basement, in a place with patchy coverage.

Costs no mobile data. Which matters enormously where data is metered and expensive.

Nothing is transmitted, so E4·L8’s paste test resolves differently. Content that could not go to a cloud service may be fine on-device.

Less capable, and adequate for a surprising amount: prediction, transcription, summarising, tidying.

Where a feature offers a choice, on-device is frequently the better default for ordinary work, with cloud reserved for things that genuinely need capability.

The costs nobody mentions

Mobile data. Cloud AI features move real amounts of data, and image and voice features move more. On a metered plan that is money — E11’s subject, arriving in a form its lessons did not cover.

Battery. On-device inference is computationally heavy. Sustained use warms the phone and drains it noticeably.

Storage. Downloaded models take gigabytes, on devices where storage is often the binding constraint.

None of these appears in the feature description. All three are real for someone using a mid-range phone on a limited plan, which is a large share of the world.

Practical settings worth checking

Three specific ones on any phone:

Which features work offline — so you know what still functions with no signal.

Whether AI features are restricted on mobile data — many can be limited to Wi-Fi, which is the single most useful setting for a metered plan.

What the assistant can access when the phone is locked, and what a voice assistant does with recordings — L3’s territory, and phones are where the defaults are most permissive.

Try it now (6 minutes)

Dictate a prompt instead of typing it — at least three sentences of context. Compare the answer to your usual shorter typed version.

Then find your phone’s setting for restricting AI features to Wi-Fi. Whether or not you enable it, knowing it exists is worth the minute.

Check your understanding

1. Voice input on a phone is often better than typing because: Speech recognition is more accurate than keyboards It is faster and you will supply far more context than you would thumb-type It uses less data

2. On-device processing means the feature: Is more capable than the cloud version Works offline, costs no data, and transmits nothing Requires a subscription

3. Which mobile cost is most often overlooked? Mobile data used by cloud AI features The subscription price Screen wear

Recap

On a phone, dictate rather than type — it is faster and you will give more context. Use the camera, asking for exact text before explanation. Prefer on-device where offered: offline, no data cost, nothing transmitted, and adequate for most ordinary work. And count the mobile-specific costs — data, battery and storage — which no feature description mentions.